| import numpy as np
|
| import matplotlib.pyplot as plt
|
| import networkx as nx
|
| from matplotlib.gridspec import GridSpec
|
| from matplotlib.patches import FancyArrowPatch
|
| from scipy.stats import norm
|
|
|
| import os
|
| import re
|
|
|
| def setup_figure(title, rows, cols):
|
| """Initializes a new figure and grid layout with constrained_layout to avoid warnings."""
|
| fig = plt.figure(figsize=(20, 10), constrained_layout=True)
|
| fig.suptitle(title, fontsize=18, fontweight='bold')
|
| gs = GridSpec(rows, cols, figure=fig)
|
| return fig, gs
|
|
|
| def plot_agent_env_loop(ax):
|
| """MDP & Environment: Agent-Environment Interaction Loop (Flowchart)."""
|
| ax.axis('off')
|
| ax.set_title("Agent-Environment Interaction", fontsize=12, fontweight='bold')
|
|
|
| props = dict(boxstyle="round,pad=0.8", fc="ivory", ec="black", lw=1.5)
|
| ax.text(0.5, 0.8, "Agent", ha="center", va="center", bbox=props, fontsize=12)
|
| ax.text(0.5, 0.2, "Environment", ha="center", va="center", bbox=props, fontsize=12)
|
|
|
|
|
|
|
| ax.annotate("Action $A_t$", xy=(0.5, 0.35), xytext=(0.5, 0.65),
|
| arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=-0.5", lw=2))
|
|
|
| ax.annotate("State $S_{t+1}$, Reward $R_{t+1}$", xy=(0.5, 0.65), xytext=(0.5, 0.35),
|
| arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=-0.5", lw=2, color='green'))
|
|
|
| def plot_mdp_graph(ax):
|
| """MDP & Environment: Directed graph with probability-weighted arrows."""
|
| G = nx.DiGraph()
|
|
|
| G.add_edges_from([
|
| ('S0', 'S1', {'weight': 0.8}), ('S0', 'S2', {'weight': 0.2}),
|
| ('S1', 'S2', {'weight': 1.0}), ('S2', 'S0', {'weight': 0.5}), ('S2', 'S2', {'weight': 0.5})
|
| ])
|
| pos = nx.spring_layout(G, seed=42)
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, node_size=1500, node_color='lightblue')
|
| nx.draw_networkx_labels(ax=ax, G=G, pos=pos, font_weight='bold')
|
|
|
| edge_labels = {(u, v): f"P={d['weight']}" for u, v, d in G.edges(data=True)}
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos, arrowsize=20, edge_color='gray', connectionstyle="arc3,rad=0.1")
|
| nx.draw_networkx_edge_labels(ax=ax, G=G, pos=pos, edge_labels=edge_labels, font_size=9)
|
| ax.set_title("MDP State Transition Graph", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_reward_landscape(fig, gs):
|
| """MDP & Environment: 3D surface plot of a reward function."""
|
|
|
| try:
|
| ax = fig.add_subplot(gs[0, 1], projection='3d')
|
| except IndexError:
|
| ax = fig.add_subplot(gs[0, 0], projection='3d')
|
| X = np.linspace(-5, 5, 50)
|
| Y = np.linspace(-5, 5, 50)
|
| X, Y = np.meshgrid(X, Y)
|
| Z = np.sin(np.sqrt(X**2 + Y**2)) + (X * 0.1)
|
|
|
| surf = ax.plot_surface(X, Y, Z, cmap='viridis', edgecolor='none', alpha=0.9)
|
| ax.set_title("Reward Function Landscape", fontsize=12, fontweight='bold')
|
| ax.set_xlabel('State X')
|
| ax.set_ylabel('State Y')
|
| ax.set_zlabel('Reward R(s)')
|
|
|
| def plot_trajectory(ax):
|
| """MDP & Environment: Trajectory / Episode Sequence."""
|
| ax.set_title("Trajectory Sequence", fontsize=12, fontweight='bold')
|
| states = ['s0', 's1', 's2', 's3', 'sT']
|
| actions = ['a0', 'a1', 'a2', 'a3']
|
| rewards = ['r1', 'r2', 'r3', 'r4']
|
|
|
| for i, s in enumerate(states):
|
| ax.text(i, 0.5, s, ha='center', va='center', bbox=dict(boxstyle="circle", fc="white"))
|
| if i < len(actions):
|
| ax.annotate("", xy=(i+0.8, 0.5), xytext=(i+0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(i+0.5, 0.6, actions[i], ha='center', color='blue')
|
| ax.text(i+0.5, 0.4, rewards[i], ha='center', color='red')
|
|
|
| ax.set_xlim(-0.5, len(states)-0.5)
|
| ax.set_ylim(0, 1)
|
| ax.axis('off')
|
|
|
| def plot_continuous_space(ax):
|
| """MDP & Environment: Continuous State/Action Space Visualization."""
|
| np.random.seed(42)
|
| x = np.random.randn(200, 2)
|
| labels = np.linalg.norm(x, axis=1) > 1.0
|
| ax.scatter(x[labels, 0], x[labels, 1], c='coral', alpha=0.6, label='High Reward')
|
| ax.scatter(x[~labels, 0], x[~labels, 1], c='skyblue', alpha=0.6, label='Low Reward')
|
| ax.set_title("Continuous State Space (2D Projection)", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_discount_decay(ax):
|
| """MDP & Environment: Discount Factor (gamma) Effect."""
|
| t = np.arange(0, 20)
|
| for gamma in [0.5, 0.9, 0.99]:
|
| ax.plot(t, gamma**t, marker='o', markersize=4, label=rf"$\gamma={gamma}$")
|
| ax.set_title(r"Discount Factor $\gamma^t$ Decay", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Time steps (t)")
|
| ax.set_ylabel("Weight")
|
| ax.legend()
|
| ax.grid(True, alpha=0.3)
|
|
|
| def plot_value_heatmap(ax):
|
| """Value & Policy: State-Value Function V(s) Heatmap (Gridworld)."""
|
| grid_size = 5
|
|
|
| values = np.zeros((grid_size, grid_size))
|
| for i in range(grid_size):
|
| for j in range(grid_size):
|
| values[i, j] = -( (grid_size-1-i)**2 + (grid_size-1-j)**2 ) * 0.5
|
| values[-1, -1] = 10.0
|
|
|
| cax = ax.matshow(values, cmap='magma')
|
| for (i, j), z in np.ndenumerate(values):
|
| ax.text(j, i, f'{z:0.1f}', ha='center', va='center', color='white' if z < -5 else 'black', fontsize=9)
|
|
|
| ax.set_title("State-Value Function V(s) Heatmap", fontsize=12, fontweight='bold', pad=15)
|
| ax.set_xticks(range(grid_size))
|
| ax.set_yticks(range(grid_size))
|
|
|
| def plot_backup_diagram(ax):
|
| """Dynamic Programming: Policy Evaluation Backup Diagram."""
|
| G = nx.DiGraph()
|
| G.add_node("s", layer=0)
|
| G.add_node("a1", layer=1); G.add_node("a2", layer=1)
|
| G.add_node("s'_1", layer=2); G.add_node("s'_2", layer=2); G.add_node("s'_3", layer=2)
|
|
|
| G.add_edges_from([("s", "a1"), ("s", "a2")])
|
| G.add_edges_from([("a1", "s'_1"), ("a1", "s'_2"), ("a2", "s'_3")])
|
|
|
| pos = {
|
| "s": (0.5, 1),
|
| "a1": (0.25, 0.5), "a2": (0.75, 0.5),
|
| "s'_1": (0.1, 0), "s'_2": (0.4, 0), "s'_3": (0.75, 0)
|
| }
|
|
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, nodelist=["s", "s'_1", "s'_2", "s'_3"], node_size=800, node_color='white', edgecolors='black')
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, nodelist=["a1", "a2"], node_size=300, node_color='black')
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos, arrows=True)
|
| nx.draw_networkx_labels(ax=ax, G=G, pos=pos, labels={"s": "s", "s'_1": "s'", "s'_2": "s'", "s'_3": "s'"}, font_size=10)
|
|
|
| ax.set_title("DP Policy Eval Backup", fontsize=12, fontweight='bold')
|
| ax.set_ylim(-0.2, 1.2)
|
| ax.axis('off')
|
|
|
| def plot_action_value_q(ax):
|
| """Value & Policy: Action-Value Function Q(s,a) (Heatmap per action stack)."""
|
| grid = np.random.rand(3, 3)
|
| ax.imshow(grid, cmap='YlGnBu')
|
| for (i, j), z in np.ndenumerate(grid):
|
| ax.text(j, i, f'{z:0.1f}', ha='center', va='center', fontsize=8)
|
| ax.set_title(r"Action-Value $Q(s, a_{up})$", fontsize=12, fontweight='bold')
|
| ax.set_xticks([]); ax.set_yticks([])
|
|
|
| def plot_policy_arrows(ax):
|
| """Value & Policy: Policy π(s) as arrow overlays on grid."""
|
| grid_size = 4
|
| ax.set_xlim(-0.5, grid_size-0.5)
|
| ax.set_ylim(-0.5, grid_size-0.5)
|
| for i in range(grid_size):
|
| for j in range(grid_size):
|
| dx, dy = np.random.choice([0, 0.3, -0.3]), np.random.choice([0, 0.3, -0.3])
|
| if dx == 0 and dy == 0: dx = 0.3
|
| ax.add_patch(FancyArrowPatch((j, i), (j+dx, i+dy), arrowstyle='->', mutation_scale=15))
|
| ax.set_title(r"Policy $\pi(s)$ Arrows", fontsize=12, fontweight='bold')
|
| ax.set_xticks(range(grid_size)); ax.set_yticks(range(grid_size)); ax.grid(True, alpha=0.2)
|
|
|
| def plot_advantage_function(ax):
|
| """Value & Policy: Advantage Function A(s,a) = Q-V."""
|
| actions = ['A1', 'A2', 'A3', 'A4']
|
| advantage = [2.1, -1.2, 0.5, -0.8]
|
| colors = ['green' if v > 0 else 'red' for v in advantage]
|
| ax.bar(actions, advantage, color=colors, alpha=0.7)
|
| ax.axhline(0, color='black', lw=1)
|
| ax.set_title(r"Advantage $A(s, a)$", fontsize=12, fontweight='bold')
|
| ax.set_ylabel("Value")
|
|
|
| def plot_policy_improvement(ax):
|
| """Dynamic Programming: Policy Improvement (Before vs After)."""
|
| ax.axis('off')
|
| ax.set_title("Policy Improvement", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"$\pi_{old}$", fontsize=15, bbox=dict(boxstyle="round", fc="lightgrey"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->", lw=2))
|
| ax.text(0.5, 0.6, "Greedy\nImprovement", ha='center', fontsize=9)
|
| ax.text(0.85, 0.5, r"$\pi_{new}$", fontsize=15, bbox=dict(boxstyle="round", fc="lightgreen"))
|
|
|
| def plot_value_iteration_backup(ax):
|
| """Dynamic Programming: Value Iteration Backup Diagram (Max over actions)."""
|
| G = nx.DiGraph()
|
| pos = {"s": (0.5, 1), "max": (0.5, 0.5), "s1": (0.2, 0), "s2": (0.5, 0), "s3": (0.8, 0)}
|
| G.add_nodes_from(pos.keys())
|
| G.add_edges_from([("s", "max"), ("max", "s1"), ("max", "s2"), ("max", "s3")])
|
|
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, node_size=500, node_color='white', edgecolors='black')
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos, arrows=True)
|
| nx.draw_networkx_labels(ax=ax, G=G, pos=pos, labels={"s": "s", "max": "max", "s1": "s'", "s2": "s'", "s3": "s'"}, font_size=9)
|
| ax.set_title("Value Iteration Backup", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_policy_iteration_cycle(ax):
|
| """Dynamic Programming: Policy Iteration Full Cycle Flowchart."""
|
| ax.axis('off')
|
| ax.set_title("Policy Iteration Cycle", fontsize=12, fontweight='bold')
|
| props = dict(boxstyle="round", fc="aliceblue", ec="black")
|
| ax.text(0.5, 0.8, r"Policy Evaluation" + "\n" + r"$V \leftarrow V^\pi$", ha="center", bbox=props)
|
| ax.text(0.5, 0.2, r"Policy Improvement" + "\n" + r"$\pi \leftarrow \text{greedy}(V)$", ha="center", bbox=props)
|
| ax.annotate("", xy=(0.7, 0.3), xytext=(0.7, 0.7), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=-0.5"))
|
| ax.annotate("", xy=(0.3, 0.7), xytext=(0.3, 0.3), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=-0.5"))
|
|
|
| def plot_mc_backup(ax):
|
| """Monte Carlo: Backup diagram (Full trajectory until terminal sT)."""
|
| ax.axis('off')
|
| ax.set_title("Monte Carlo Backup", fontsize=12, fontweight='bold')
|
| nodes = ['s', 's1', 's2', 'sT']
|
| pos = {n: (0.5, 0.9 - i*0.25) for i, n in enumerate(nodes)}
|
| for i in range(len(nodes)-1):
|
| ax.annotate("", xy=pos[nodes[i+1]], xytext=pos[nodes[i]], arrowprops=dict(arrowstyle="->", lw=1.5))
|
| ax.text(pos[nodes[i]][0]+0.05, pos[nodes[i]][1], nodes[i], va='center')
|
| ax.text(pos['sT'][0]+0.05, pos['sT'][1], 'sT', va='center', fontweight='bold')
|
| ax.annotate("Update V(s) using G", xy=(0.3, 0.9), xytext=(0.3, 0.15), arrowprops=dict(arrowstyle="->", color='red', connectionstyle="arc3,rad=0.3"))
|
|
|
| def plot_mcts(ax):
|
| """Monte Carlo: Monte Carlo Tree Search (MCTS) tree diagram."""
|
| G = nx.balanced_tree(2, 2, create_using=nx.DiGraph())
|
| pos = nx.drawing.nx_agraph.graphviz_layout(G, prog='dot') if 'pygraphviz' in globals() else nx.shell_layout(G)
|
|
|
| pos = {0:(0,0), 1:(-1,-1), 2:(1,-1), 3:(-1.5,-2), 4:(-0.5,-2), 5:(0.5,-2), 6:(1.5,-2)}
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, node_size=300, node_color='lightyellow', edgecolors='black')
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos, arrows=True)
|
| ax.set_title("MCTS Tree", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_importance_sampling(ax):
|
| """Monte Carlo: Importance Sampling Ratio Flow."""
|
| ax.axis('off')
|
| ax.set_title("Importance Sampling", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"$\pi(a|s)$", bbox=dict(boxstyle="circle", fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.2, r"$b(a|s)$", bbox=dict(boxstyle="circle", fc="lightpink"), ha='center')
|
| ax.annotate(r"$\rho = \frac{\pi}{b}$", xy=(0.7, 0.5), fontsize=15)
|
| ax.annotate("", xy=(0.5, 0.35), xytext=(0.5, 0.65), arrowprops=dict(arrowstyle="<->", lw=2))
|
|
|
| def plot_td_backup(ax):
|
| """Temporal Difference: TD(0) 1-step backup."""
|
| ax.axis('off')
|
| ax.set_title("TD(0) Backup", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "s", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.5, 0.2, "s'", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.annotate(r"$R + \gamma V(s')$", xy=(0.5, 0.4), ha='center', color='blue')
|
| ax.annotate("", xy=(0.5, 0.35), xytext=(0.5, 0.65), arrowprops=dict(arrowstyle="<-", lw=2))
|
|
|
| def plot_nstep_td(ax):
|
| """Temporal Difference: n-step TD backup."""
|
| ax.axis('off')
|
| ax.set_title("n-step TD Backup", fontsize=12, fontweight='bold')
|
| for i in range(4):
|
| ax.text(0.5, 0.9-i*0.2, f"s_{i}", bbox=dict(boxstyle="circle", fc="white"), ha='center', fontsize=8)
|
| if i < 3: ax.annotate("", xy=(0.5, 0.75-i*0.2), xytext=(0.5, 0.85-i*0.2), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate(r"$G_t^{(n)}$", xy=(0.7, 0.5), fontsize=12, color='red')
|
|
|
| def plot_eligibility_traces(ax):
|
| """Temporal Difference: TD(lambda) Eligibility Traces decay curve."""
|
| t = np.arange(0, 50)
|
|
|
| trace = np.zeros_like(t, dtype=float)
|
| visits = [5, 20, 35]
|
| for v in visits:
|
| trace[v:] += (0.8 ** np.arange(len(t)-v))
|
| ax.plot(t, trace, color='brown', lw=2)
|
| ax.set_title(r"Eligibility Trace $z_t(\lambda)$", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Time")
|
| ax.fill_between(t, trace, color='brown', alpha=0.1)
|
|
|
| def plot_sarsa_backup(ax):
|
| """Temporal Difference: SARSA (On-policy) backup."""
|
| ax.axis('off')
|
| ax.set_title("SARSA Backup", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "(s,a)", ha='center')
|
| ax.text(0.5, 0.1, "(s',a')", ha='center')
|
| ax.annotate("", xy=(0.5, 0.2), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="<-", lw=2, color='orange'))
|
| ax.text(0.6, 0.5, "On-policy", rotation=90)
|
|
|
| def plot_q_learning_backup(ax):
|
| """Temporal Difference: Q-Learning (Off-policy) backup."""
|
| ax.axis('off')
|
| ax.set_title("Q-Learning Backup", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "(s,a)", ha='center')
|
| ax.text(0.5, 0.1, r"$\max_{a'} Q(s',a')$", ha='center', bbox=dict(boxstyle="round", fc="lightcyan"))
|
| ax.annotate("", xy=(0.5, 0.25), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="<-", lw=2, color='blue'))
|
|
|
| def plot_double_q(ax):
|
| """Temporal Difference: Double Q-Learning / Double DQN."""
|
| ax.axis('off')
|
| ax.set_title("Double Q-Learning", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Network A", bbox=dict(fc="lightyellow"), ha='center')
|
| ax.text(0.5, 0.2, "Network B", bbox=dict(fc="lightcyan"), ha='center')
|
| ax.annotate("Select $a^*$", xy=(0.3, 0.8), xytext=(0.5, 0.85), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Eval $Q(s', a^*)$", xy=(0.7, 0.2), xytext=(0.5, 0.15), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_dueling_dqn(ax):
|
| """Temporal Difference: Dueling DQN Architecture."""
|
| ax.axis('off')
|
| ax.set_title("Dueling DQN", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Backbone", bbox=dict(fc="lightgrey"), ha='center', rotation=90)
|
| ax.text(0.5, 0.7, "V(s)", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.3, "A(s,a)", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.9, 0.5, "Q(s,a)", bbox=dict(boxstyle="circle", fc="orange"), ha='center')
|
| ax.annotate("", xy=(0.35, 0.7), xytext=(0.15, 0.55), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.35, 0.3), xytext=(0.15, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.55), xytext=(0.6, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.45), xytext=(0.6, 0.3), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_prioritized_replay(ax):
|
| """Temporal Difference: Prioritized Experience Replay (PER)."""
|
| priorities = np.random.pareto(3, 100)
|
| ax.hist(priorities, bins=20, color='teal', alpha=0.7)
|
| ax.set_title("Prioritized Replay (TD-Error)", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Priority $P_i$")
|
| ax.set_ylabel("Count")
|
|
|
| def plot_rainbow_dqn(ax):
|
| """Temporal Difference: Rainbow DQN Composite."""
|
| ax.axis('off')
|
| ax.set_title("Rainbow DQN", fontsize=12, fontweight='bold')
|
| features = ["Double", "Dueling", "PER", "Noisy", "Distributional", "n-step"]
|
| for i, f in enumerate(features):
|
| ax.text(0.5, 0.9 - i*0.15, f, ha='center', bbox=dict(boxstyle="round", fc="ghostwhite"), fontsize=8)
|
|
|
| def plot_linear_fa(ax):
|
| """Function Approximation: Linear Function Approximation."""
|
| ax.axis('off')
|
| ax.set_title("Linear Function Approx", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"$\phi(s)$ Features", ha='center', bbox=dict(fc="white"))
|
| ax.text(0.5, 0.2, r"$w^T \phi(s)$", ha='center', bbox=dict(fc="lightgrey"))
|
| ax.annotate("", xy=(0.5, 0.35), xytext=(0.5, 0.65), arrowprops=dict(arrowstyle="->", lw=2))
|
|
|
| def plot_nn_layers(ax):
|
| """Function Approximation: Neural Network Layers diagram."""
|
| ax.axis('off')
|
| ax.set_title("NN Layers (Deep RL)", fontsize=12, fontweight='bold')
|
| layers = [4, 8, 8, 2]
|
| for i, l in enumerate(layers):
|
| for j in range(l):
|
| ax.scatter(i*0.3, j*0.1 - l*0.05, s=20, c='black')
|
| ax.set_xlim(-0.1, 1.0)
|
| ax.set_ylim(-0.5, 0.5)
|
|
|
| def plot_computation_graph(ax):
|
| """Function Approximation: Computation Graph / Backprop Flow."""
|
| ax.axis('off')
|
| ax.set_title("Computation Graph (DAG)", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Input", bbox=dict(boxstyle="circle", fc="white"))
|
| ax.text(0.5, 0.5, "Op", bbox=dict(boxstyle="square", fc="lightgrey"))
|
| ax.text(0.9, 0.5, "Loss", bbox=dict(boxstyle="circle", fc="salmon"))
|
| ax.annotate("", xy=(0.35, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Grad", xy=(0.1, 0.3), xytext=(0.9, 0.3), arrowprops=dict(arrowstyle="->", color='red', connectionstyle="arc3,rad=0.2"))
|
|
|
| def plot_target_network(ax):
|
| """Function Approximation: Target Network concept."""
|
| ax.axis('off')
|
| ax.set_title("Target Network Updates", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.8, r"$Q_\theta$ (Active)", bbox=dict(fc="lightgreen"))
|
| ax.text(0.7, 0.8, r"$Q_{\theta^-}$ (Target)", bbox=dict(fc="lightblue"))
|
| ax.annotate("periodic copy", xy=(0.6, 0.8), xytext=(0.4, 0.8), arrowprops=dict(arrowstyle="<-", ls='--'))
|
|
|
| def plot_ppo_clip(ax):
|
| """Policy Gradients: PPO Clipped Surrogate Objective."""
|
| epsilon = 0.2
|
| r = np.linspace(0.5, 1.5, 100)
|
| advantage = 1.0
|
| surr1 = r * advantage
|
| surr2 = np.clip(r, 1-epsilon, 1+epsilon) * advantage
|
| ax.plot(r, surr1, '--', label="r*A")
|
| ax.plot(r, np.minimum(surr1, surr2), 'r', label="min(r*A, clip*A)")
|
| ax.set_title("PPO-Clip Objective", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
| ax.axvline(1, color='gray', linestyle=':')
|
|
|
| def plot_trpo_trust_region(ax):
|
| """Policy Gradients: TRPO Trust Region / KL Constraint."""
|
| ax.set_title("TRPO Trust Region", fontsize=12, fontweight='bold')
|
| circle = plt.Circle((0.5, 0.5), 0.3, color='blue', fill=False, label="KL Constraint")
|
| ax.add_artist(circle)
|
| ax.scatter(0.5, 0.5, c='black', label=r"$\pi_{old}$")
|
| ax.arrow(0.5, 0.5, 0.15, 0.1, head_width=0.03, color='red', label="Update")
|
| ax.set_xlim(0, 1); ax.set_ylim(0, 1)
|
| ax.axis('off')
|
|
|
| def plot_a3c_multi_worker(ax):
|
| """Actor-Critic: Asynchronous Multi-worker (A3C)."""
|
| ax.axis('off')
|
| ax.set_title("A3C Multi-worker", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Global Parameters", bbox=dict(fc="gold"), ha='center')
|
| for i in range(3):
|
| ax.text(0.2 + i*0.3, 0.2, f"Worker {i+1}", bbox=dict(fc="lightgrey"), ha='center', fontsize=8)
|
| ax.annotate("", xy=(0.5, 0.7), xytext=(0.2 + i*0.3, 0.3), arrowprops=dict(arrowstyle="<->"))
|
|
|
| def plot_sac_arch(ax):
|
| """Actor-Critic: SAC (Entropy-regularized)."""
|
| ax.axis('off')
|
| ax.set_title("SAC Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.7, "Actor", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.3, "Entropy Bonus", bbox=dict(fc="salmon"), ha='center')
|
| ax.text(0.1, 0.5, "State", ha='center')
|
| ax.text(0.9, 0.5, "Action", ha='center')
|
| ax.annotate("", xy=(0.4, 0.7), xytext=(0.15, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.55), xytext=(0.5, 0.4), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.85, 0.5), xytext=(0.6, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_softmax_exploration(ax):
|
| """Exploration: Softmax / Boltzmann probabilities."""
|
| x = np.arange(4)
|
| logits = [1, 2, 5, 3]
|
| for tau in [0.5, 1.0, 5.0]:
|
| probs = np.exp(np.array(logits)/tau)
|
| probs /= probs.sum()
|
| ax.plot(x, probs, marker='o', label=rf"$\tau={tau}$")
|
| ax.set_title("Softmax Exploration", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
| ax.set_xticks(x)
|
|
|
| def plot_ucb_confidence(ax):
|
| """Exploration: Upper Confidence Bound (UCB)."""
|
| actions = ['A1', 'A2', 'A3']
|
| means = [0.6, 0.8, 0.5]
|
| conf = [0.3, 0.1, 0.4]
|
| ax.bar(actions, means, yerr=conf, capsize=10, color='skyblue', label='Mean Q')
|
| ax.set_title("UCB Action Values", fontsize=12, fontweight='bold')
|
| ax.set_ylim(0, 1.2)
|
|
|
| def plot_intrinsic_motivation(ax):
|
| """Exploration: Intrinsic Motivation / Curiosity."""
|
| ax.axis('off')
|
| ax.set_title("Intrinsic Motivation", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.5, "World Model", bbox=dict(fc="lightyellow"), ha='center')
|
| ax.text(0.7, 0.5, "Prediction\nError", bbox=dict(boxstyle="circle", fc="orange"), ha='center')
|
| ax.annotate("", xy=(0.58, 0.5), xytext=(0.42, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.85, 0.5, r"$R_{int}$", fontweight='bold')
|
|
|
| def plot_entropy_bonus(ax):
|
| """Exploration: Entropy Regularization curve."""
|
| p = np.linspace(0.01, 0.99, 50)
|
| entropy = -(p * np.log(p) + (1-p) * np.log(1-p))
|
| ax.plot(p, entropy, color='purple')
|
| ax.set_title(r"Entropy $H(\pi)$", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("$P(a)$")
|
|
|
| def plot_options_framework(ax):
|
| """Hierarchical RL: Options Framework."""
|
| ax.axis('off')
|
| ax.set_title("Options Framework", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"High-level policy" + "\n" + r"$\pi_{hi}$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.2, 0.2, "Option 1", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.8, 0.2, "Option 2", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.3, 0.3), xytext=(0.45, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.3), xytext=(0.55, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_feudal_networks(ax):
|
| """Hierarchical RL: Feudal Networks / Hierarchy."""
|
| ax.axis('off')
|
| ax.set_title("Feudal Networks", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.85, "Manager", bbox=dict(fc="plum"), ha='center')
|
| ax.text(0.5, 0.15, "Worker", bbox=dict(fc="wheat"), ha='center')
|
| ax.annotate("Goal $g_t$", xy=(0.5, 0.3), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->", lw=2))
|
|
|
| def plot_world_model(ax):
|
| """Model-Based RL: Learned Dynamics Model."""
|
| ax.axis('off')
|
| ax.set_title("World Model (Dynamics)", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "(s,a)", ha='center')
|
| ax.text(0.5, 0.5, r"$\hat{P}$", bbox=dict(boxstyle="circle", fc="lightgrey"), ha='center')
|
| ax.text(0.9, 0.7, r"$\hat{s}'$", ha='center')
|
| ax.text(0.9, 0.3, r"$\hat{r}$", ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.65), xytext=(0.6, 0.55), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.35), xytext=(0.6, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_model_planning(ax):
|
| """Model-Based RL: Planning / Rollouts in imagination."""
|
| ax.axis('off')
|
| ax.set_title("Model-Based Planning", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Real s", ha='center', fontweight='bold')
|
| for i in range(3):
|
| ax.annotate("", xy=(0.3+i*0.2, 0.5+(i%2)*0.1), xytext=(0.1+i*0.2, 0.5), arrowprops=dict(arrowstyle="->", color='gray'))
|
| ax.text(0.3+i*0.2, 0.55+(i%2)*0.1, "imagined", fontsize=7)
|
|
|
| def plot_offline_rl(ax):
|
| """Offline RL: Fixed dataset of trajectories."""
|
| ax.axis('off')
|
| ax.set_title("Offline RL Dataset", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.5, r"Static" + "\n" + r"Dataset" + "\n" + r"$\mathcal{D}$", bbox=dict(boxstyle="round", fc="lightgrey"), ha='center')
|
| ax.annotate("No interaction", xy=(0.5, 0.9), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->", color='red'))
|
| ax.scatter([0.2, 0.8, 0.3, 0.7], [0.8, 0.8, 0.2, 0.2], marker='x', color='blue')
|
|
|
| def plot_cql_regularization(ax):
|
| """Offline RL: CQL regularization visualization."""
|
| q = np.linspace(-5, 5, 100)
|
| penalty = q**2 * 0.1
|
| ax.plot(q, penalty, 'r', label='CQL Penalty')
|
| ax.set_title("CQL Regularization", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Q-value")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_multi_agent_interaction(ax):
|
| """Multi-Agent RL: Agents communicating or competing."""
|
| G = nx.complete_graph(3)
|
| pos = nx.spring_layout(G)
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, node_size=500, node_color=['red', 'blue', 'green'])
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos, style='dashed')
|
| ax.set_title("Multi-Agent Interaction", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_ctde(ax):
|
| """Multi-Agent RL: Centralized Training Decentralized Execution (CTDE)."""
|
| ax.axis('off')
|
| ax.set_title("CTDE Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Centralized Critic", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.2, 0.2, "Agent 1", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.2, "Agent 2", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.7), xytext=(0.25, 0.35), arrowprops=dict(arrowstyle="<-", color='gray'))
|
| ax.annotate("", xy=(0.5, 0.7), xytext=(0.75, 0.35), arrowprops=dict(arrowstyle="<-", color='gray'))
|
|
|
| def plot_payoff_matrix(ax):
|
| """Multi-Agent RL: Cooperative / Competitive Payoff Matrix."""
|
| matrix = np.array([[(3,3), (0,5)], [(5,0), (1,1)]])
|
| ax.axis('off')
|
| ax.set_title("Payoff Matrix (Prisoner's)", fontsize=12, fontweight='bold')
|
| for i in range(2):
|
| for j in range(2):
|
| ax.text(j, 1-i, str(matrix[i, j]), ha='center', va='center', bbox=dict(fc="white"))
|
| ax.set_xlim(-0.5, 1.5); ax.set_ylim(-0.5, 1.5)
|
|
|
| def plot_irl_reward_inference(ax):
|
| """Inverse RL: Infer reward from expert demonstrations."""
|
| ax.axis('off')
|
| ax.set_title("Inferred Reward Heatmap", fontsize=12, fontweight='bold')
|
| grid = np.zeros((5, 5))
|
| grid[2:4, 2:4] = 1.0
|
| ax.imshow(grid, cmap='hot')
|
|
|
| def plot_gail_flow(ax):
|
| """Inverse RL: GAIL (Generative Adversarial Imitation Learning)."""
|
| ax.axis('off')
|
| ax.set_title("GAIL Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.8, "Expert Data", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.2, 0.2, "Policy (Gen)", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.8, 0.5, "Discriminator", bbox=dict(boxstyle="square", fc="salmon"), ha='center')
|
| ax.annotate("", xy=(0.6, 0.55), xytext=(0.35, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.6, 0.45), xytext=(0.35, 0.25), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_meta_rl_nested_loop(ax):
|
| """Meta-RL: Outer loop (meta) + inner loop (adaptation)."""
|
| ax.axis('off')
|
| ax.set_title("Meta-RL Loops", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.4, fill=False, ls='--'))
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.2, fill=False))
|
| ax.text(0.5, 0.5, "Inner\nLoop", ha='center', fontsize=8)
|
| ax.text(0.5, 0.8, "Outer Loop", ha='center', fontsize=10)
|
|
|
| def plot_task_distribution(ax):
|
| """Meta-RL: Multiple MDPs from distribution."""
|
| ax.axis('off')
|
| ax.set_title("Task Distribution", fontsize=12, fontweight='bold')
|
| for i in range(3):
|
| ax.text(0.2 + i*0.3, 0.5, f"Task {i+1}", bbox=dict(boxstyle="round", fc="ivory"), fontsize=8)
|
| ax.annotate("sample", xy=(0.5, 0.8), xytext=(0.5, 0.6), arrowprops=dict(arrowstyle="<-"))
|
|
|
| def plot_replay_buffer(ax):
|
| """Advanced: Experience Replay Buffer (FIFO)."""
|
| ax.axis('off')
|
| ax.set_title("Experience Replay Buffer", fontsize=12, fontweight='bold')
|
| for i in range(5):
|
| ax.add_patch(plt.Rectangle((0.1+i*0.15, 0.4), 0.1, 0.2, fill=True, color='lightgrey'))
|
| ax.text(0.15+i*0.15, 0.5, f"e_{i}", ha='center')
|
| ax.annotate("In", xy=(0.05, 0.5), xytext=(-0.1, 0.5), arrowprops=dict(arrowstyle="->"), annotation_clip=False)
|
| ax.annotate("Out (Batch)", xy=(0.85, 0.5), xytext=(1.0, 0.5), arrowprops=dict(arrowstyle="<-"), annotation_clip=False)
|
|
|
| def plot_state_visitation(ax):
|
| """Advanced: State Visitation / Occupancy Measure."""
|
| data = np.random.multivariate_normal([0, 0], [[1, 0.5], [0.5, 1]], 1000)
|
| ax.hexbin(data[:, 0], data[:, 1], gridsize=15, cmap='Blues')
|
| ax.set_title("State Visitation Heatmap", fontsize=12, fontweight='bold')
|
|
|
| def plot_regret_curve(ax):
|
| """Advanced: Regret / Cumulative Regret."""
|
| t = np.arange(100)
|
| regret = np.sqrt(t) + np.random.normal(0, 0.5, 100)
|
| ax.plot(t, regret, color='red', label='Sub-linear Regret')
|
| ax.set_title("Cumulative Regret", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Time")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_attention_weights(ax):
|
| """Advanced: Attention Mechanisms (Heatmap)."""
|
| weights = np.random.rand(5, 5)
|
| ax.imshow(weights, cmap='viridis')
|
| ax.set_title("Attention Weight Matrix", fontsize=12, fontweight='bold')
|
| ax.set_xticks([]); ax.set_yticks([])
|
|
|
| def plot_diffusion_policy(ax):
|
| """Advanced: Diffusion Policy denoising steps."""
|
| ax.axis('off')
|
| ax.set_title("Diffusion Policy (Denoising)", fontsize=12, fontweight='bold')
|
| for i in range(4):
|
| ax.scatter(0.1+i*0.25, 0.5, s=100/(i+1), c='black', alpha=1.0 - i*0.2)
|
| if i < 3: ax.annotate("", xy=(0.25+i*0.25, 0.5), xytext=(0.15+i*0.25, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.3, "Noise $\\rightarrow$ Action", ha='center', fontsize=8)
|
|
|
| def plot_gnn_rl(ax):
|
| """Advanced: Graph Neural Networks for RL."""
|
| G = nx.star_graph(4)
|
| pos = nx.spring_layout(G)
|
| nx.draw_networkx_nodes(ax=ax, G=G, pos=pos, node_size=200, node_color='orange')
|
| nx.draw_networkx_edges(ax=ax, G=G, pos=pos)
|
| ax.set_title("GNN Message Passing", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_latent_space(ax):
|
| """Advanced: World Model / Latent Space."""
|
| ax.axis('off')
|
| ax.set_title("Latent Space (VAE/Dreamer)", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Image", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.5, 0.5, "Latent $z$", bbox=dict(boxstyle="circle", fc="lightpink"), ha='center')
|
| ax.text(0.9, 0.5, "Reconstruction", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_convergence_log(ax):
|
| """Advanced: Convergence Analysis Plots (Log-scale)."""
|
| iterations = np.arange(1, 100)
|
| error = 10 / iterations**2
|
| ax.loglog(iterations, error, color='green')
|
| ax.set_title("Value Convergence (Log)", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Iterations")
|
| ax.set_ylabel("Error")
|
| ax.grid(True, which="both", ls="-", alpha=0.3)
|
|
|
| def plot_expected_sarsa_backup(ax):
|
| """Temporal Difference: Expected SARSA (Expectation over policy)."""
|
| ax.axis('off')
|
| ax.set_title("Expected SARSA Backup", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "(s,a)", ha='center')
|
| ax.text(0.5, 0.1, r"$\sum_{a'} \pi(a'|s') Q(s',a')$", ha='center', bbox=dict(boxstyle="round", fc="ivory"))
|
| ax.annotate("", xy=(0.5, 0.25), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="<-", lw=2, color='purple'))
|
|
|
| def plot_reinforce_flow(ax):
|
| """Policy Gradients: REINFORCE (Full trajectory flow)."""
|
| ax.axis('off')
|
| ax.set_title("REINFORCE Flow", fontsize=12, fontweight='bold')
|
| steps = ["s0", "a0", "r1", "s1", "...", "GT"]
|
| for i, s in enumerate(steps):
|
| ax.text(0.1 + i*0.15, 0.5, s, bbox=dict(boxstyle="circle", fc="white"))
|
| ax.annotate(r"$\nabla_\theta J \propto G_t \nabla \ln \pi$", xy=(0.5, 0.8), ha='center', fontsize=12, color='darkgreen')
|
|
|
| def plot_advantage_scaled_grad(ax):
|
| """Policy Gradients: Baseline / Advantage scaled gradient."""
|
| ax.axis('off')
|
| ax.set_title("Baseline Subtraction", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"$(G_t - b(s))$", bbox=dict(fc="salmon"), ha='center')
|
| ax.text(0.5, 0.3, r"Scale $\nabla \ln \pi$", ha='center')
|
| ax.annotate("", xy=(0.5, 0.4), xytext=(0.5, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_skill_discovery(ax):
|
| """Hierarchical RL: Skill Discovery (Unsupervised clusters)."""
|
| np.random.seed(0)
|
| for i in range(3):
|
| center = np.random.randn(2) * 2
|
| pts = np.random.randn(20, 2) * 0.5 + center
|
| ax.scatter(pts[:, 0], pts[:, 1], alpha=0.6, label=f"Skill {i+1}")
|
| ax.set_title("Skill Embedding Space", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_imagination_rollout(ax):
|
| """Model-Based RL: Imagination-Augmented Rollouts (I2A)."""
|
| ax.axis('off')
|
| ax.set_title("Imagination Rollout (I2A)", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Input s", ha='center')
|
| ax.add_patch(plt.Rectangle((0.3, 0.3), 0.4, 0.4, fill=True, color='lavender'))
|
| ax.text(0.5, 0.5, "Imagination\nModule", ha='center')
|
| ax.annotate("Imagined Paths", xy=(0.8, 0.5), xytext=(0.5, 0.5), arrowprops=dict(arrowstyle="->", color='gray', connectionstyle="arc3,rad=0.3"))
|
|
|
| def plot_policy_gradient_flow(ax):
|
| """Policy Gradients: Gradient flow from reward to log-prob (DAG)."""
|
| ax.axis('off')
|
| ax.set_title("Policy Gradient Flow (DAG)", fontsize=12, fontweight='bold')
|
|
|
| bbox_props = dict(boxstyle="round,pad=0.5", fc="lightgrey", ec="black", lw=1.5)
|
| ax.text(0.1, 0.8, r"Trajectory $\tau$", ha="center", va="center", bbox=bbox_props)
|
| ax.text(0.5, 0.8, r"Reward $R(\tau)$", ha="center", va="center", bbox=bbox_props)
|
| ax.text(0.1, 0.2, r"Log-Prob $\log \pi_\theta$", ha="center", va="center", bbox=bbox_props)
|
| ax.text(0.7, 0.5, r"$\nabla_\theta J(\theta)$", ha="center", va="center", bbox=dict(boxstyle="circle,pad=0.3", fc="gold", ec="black"))
|
|
|
|
|
| ax.annotate("", xy=(0.35, 0.8), xytext=(0.2, 0.8), arrowprops=dict(arrowstyle="->", lw=2))
|
| ax.annotate("", xy=(0.7, 0.65), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->", lw=2))
|
| ax.annotate("", xy=(0.6, 0.4), xytext=(0.25, 0.2), arrowprops=dict(arrowstyle="->", lw=2))
|
|
|
| def plot_rl_as_inference_pgm(ax):
|
| """PGM: RL as Inference (Control as Inference)."""
|
| ax.axis('off')
|
| ax.set_title("RL as Inference (PGM)", fontsize=12, fontweight='bold')
|
| nodes = {
|
| 's_t': (0.1, 0.8), 'a_t': (0.1, 0.4), 's_tp1': (0.5, 0.8),
|
| 'r_t': (0.5, 0.4), 'O_t': (0.8, 0.4)
|
| }
|
| for name, pos in nodes.items():
|
| color = 'white' if 'O' not in name else 'lightcoral'
|
| ax.text(pos[0], pos[1], name, bbox=dict(boxstyle="circle", fc=color), ha='center')
|
|
|
|
|
| arrows = [('s_t', 's_tp1'), ('a_t', 's_tp1'), ('s_t', 'a_t'), ('a_t', 'r_t'), ('r_t', 'O_t')]
|
| for start, end in arrows:
|
| ax.annotate("", xy=nodes[end], xytext=nodes[start], arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_rl_taxonomy_tree(ax):
|
| """Taxonomy: RL Algorithm Classification Tree."""
|
| ax.axis('off')
|
| ax.set_title("RL Algorithm Taxonomy", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "Reinforcement Learning", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.25, 0.6, "Model-Free", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.75, 0.6, "Model-Based", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.1, 0.3, "Policy Opt", fontsize=8, ha='center')
|
| ax.text(0.4, 0.3, "Value-Based", fontsize=8, ha='center')
|
| for x in [0.25, 0.75]: ax.annotate("", xy=(x, 0.65), xytext=(0.5, 0.85), arrowprops=dict(arrowstyle="->"))
|
| for x in [0.1, 0.4]: ax.annotate("", xy=(x, 0.35), xytext=(0.25, 0.55), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_distributional_rl_atoms(ax):
|
| """Distributional RL: C51 return probability atoms."""
|
| returns = np.linspace(-10, 10, 51)
|
| probs = np.exp(-(returns - 2)**2 / 4) + np.exp(-(returns + 4)**2 / 2)
|
| probs /= probs.sum()
|
| ax.bar(returns, probs, width=0.3, color='steelblue', alpha=0.8)
|
| ax.set_title("Distributional RL (Atoms)", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Return $Z$")
|
| ax.set_ylabel("Probability")
|
|
|
| def plot_her_goal_relabeling(ax):
|
| """HER: Hindsight Experience Replay goal relabeling."""
|
| ax.axis('off')
|
| ax.set_title("HER Goal Relabeling", fontsize=12, fontweight='bold')
|
| path = np.array([[0.1, 0.2], [0.3, 0.4], [0.6, 0.5], [0.8, 0.7]])
|
| ax.plot(path[:, 0], path[:, 1], 'k--', alpha=0.3)
|
| ax.scatter(path[:, 0], path[:, 1], c='black', s=20)
|
| ax.text(0.9, 0.9, "True Goal G", color='red', fontweight='bold', ha='center')
|
| ax.text(0.8, 0.6, "Relabeled G'", color='blue', fontweight='bold', ha='center')
|
| ax.annotate("", xy=(0.8, 0.7), xytext=(0.8, 0.63), arrowprops=dict(arrowstyle="->", color='blue'))
|
|
|
| def plot_dyna_q_flow(ax):
|
| """Dyna-Q: Real interaction + Model-based planning flow."""
|
| ax.axis('off')
|
| ax.set_title("Dyna-Q Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Agent Policy", bbox=dict(fc="white"), ha='center')
|
| ax.text(0.2, 0.5, "Real World", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.8, 0.5, "Model", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.5, 0.2, "Value Function / Q", bbox=dict(fc="gold"), ha='center')
|
|
|
| ax.annotate("Direct RL", xy=(0.35, 0.25), xytext=(0.2, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Planning", xy=(0.65, 0.25), xytext=(0.8, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_noisy_nets_parameters(ax):
|
| """Noisy Nets: Parameter noise distribution σ for weights."""
|
| x = np.linspace(-3, 3, 100)
|
| y = np.exp(-x**2 / 2)
|
| ax.plot(x, y, color='black', label=r"$\mu$ (Mean)")
|
| ax.fill_between(x, y-0.2, y+0.2, color='gray', alpha=0.3, label=r"$\sigma \cdot \epsilon$ (Noise)")
|
| ax.set_title("Noisy Nets Parameter Noise", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_icm_curiosity(ax):
|
| """Exploration: Intrinsic Curiosity Module (ICM)."""
|
| ax.axis('off')
|
| ax.set_title("ICM: Inverse & Forward Models", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "s_t, s_t+1", ha='center')
|
| ax.text(0.5, 0.8, "Inverse Model", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.2, "Forward Model", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.9, 0.5, "Intrinsic Reward", ha='center', color='red')
|
| ax.annotate("", xy=(0.35, 0.75), xytext=(0.2, 0.55), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.35, 0.25), xytext=(0.2, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.65, 0.3), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_v_trace_impala(ax):
|
| """IMPALA: V-trace asynchronous importance sampling."""
|
| ax.axis('off')
|
| ax.set_title("V-trace (IMPALA)", fontsize=12, fontweight='bold')
|
| for i in range(4):
|
| h = 0.5 + 0.3*np.sin(i)
|
| ax.bar(0.2+i*0.2, h, width=0.1, color='teal')
|
| ax.text(0.2+i*0.2, h+0.05, rf"$\rho_{i}$", ha='center', fontsize=8)
|
| ax.axhline(0.5, ls='--', color='red', label="Clipped $\\rho$")
|
| ax.set_ylim(0, 1.2)
|
|
|
| def plot_qmix_mixing_net(ax):
|
| """Multi-Agent RL: QMIX Mixing Network."""
|
| ax.axis('off')
|
| ax.set_title("QMIX Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Mixing Network", bbox=dict(boxstyle="round,pad=1", fc="gold"), ha='center')
|
| for i in range(3):
|
| ax.text(0.2+i*0.3, 0.4, f"Agent {i+1} Q", bbox=dict(fc="grey"), ha='center', fontsize=7)
|
| ax.annotate("", xy=(0.5, 0.65), xytext=(0.2+i*0.3, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.1, "Global State s", ha='center')
|
| ax.annotate("hypernets", xy=(0.5, 0.68), xytext=(0.5, 0.2), arrowprops=dict(arrowstyle="->", ls=':'))
|
|
|
| def plot_saliency_heatmaps(ax):
|
| """Interpretability: Attention/Saliency Heatmap on input."""
|
|
|
| img = np.zeros((20, 20))
|
| img[15, 8:12] = 1.0
|
| img[5:7, 5:15] = 0.5
|
| heatmap = np.random.rand(20, 20) * 0.5
|
| heatmap[14:17, 7:13] = 1.0
|
| ax.imshow(img, cmap='gray')
|
| ax.imshow(heatmap, cmap='hot', alpha=0.5)
|
| ax.set_title("Action Saliency Heatmap", fontsize=12, fontweight='bold')
|
| ax.axis('off')
|
|
|
| def plot_action_selection_noise(ax):
|
| """Exploration: OU-noise vs Gaussian Noise paths."""
|
| t = np.arange(100)
|
| gaussian = np.random.normal(0, 0.1, 100)
|
| ou = np.zeros(100)
|
| for i in range(1, 100):
|
| ou[i] = ou[i-1] * 0.9 + np.random.normal(0, 0.1)
|
| ax.plot(t, gaussian, label="Gaussian", alpha=0.5)
|
| ax.plot(t, ou, label="Ornstein-Uhlenbeck", color='red')
|
| ax.set_title("Action Selection Noise", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_tsne_state_embeddings(ax):
|
| """Interpretability: t-SNE / UMAP State Clusters."""
|
| np.random.seed(42)
|
| for i in range(3):
|
| center = np.random.randn(2) * 5
|
| pts = np.random.randn(30, 2) + center
|
| ax.scatter(pts[:, 0], pts[:, 1], alpha=0.6, label=f"Cluster {i+1}")
|
| ax.set_title("t-SNE State Embeddings", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_loss_landscape(fig, gs):
|
| """Optimization: Loss Landscape / Surface."""
|
| ax = fig.add_subplot(gs[0, 0], projection='3d')
|
| x = np.linspace(-2, 2, 30)
|
| y = np.linspace(-2, 2, 30)
|
| X, Y = np.meshgrid(x, y)
|
| Z = X**2 + Y**2 + 0.5*np.sin(5*X)
|
| ax.plot_surface(X, Y, Z, cmap='terrain', alpha=0.8)
|
| ax.set_title("Policy Loss Landscape", fontsize=12, fontweight='bold')
|
|
|
| def plot_success_rate_curve(ax):
|
| """Evaluation: Success Rate over training."""
|
| steps = np.linspace(0, 1e6, 100)
|
| success = 1.0 / (1.0 + np.exp(-1e-5 * (steps - 4e5)))
|
| ax.plot(steps, success, color='darkgreen', lw=2)
|
| ax.set_title("Success Rate vs Steps", fontsize=12, fontweight='bold')
|
| ax.set_ylim(-0.05, 1.05)
|
| ax.grid(True, alpha=0.3)
|
|
|
| def plot_hyperparameter_sensitivity(ax):
|
| """Analysis: Hyperparameter Sensitivity Heatmap."""
|
| lr = [1e-5, 1e-4, 1e-3]
|
| batches = [32, 64, 128]
|
| data = np.array([[60, 85, 40], [75, 95, 80], [30, 50, 45]])
|
| im = ax.imshow(data, cmap='RdYlGn')
|
| ax.set_xticks(range(3)); ax.set_xticklabels(batches)
|
| ax.set_yticks(range(3)); ax.set_yticklabels(lr)
|
| ax.set_xlabel("Batch Size"); ax.set_ylabel("Learning Rate")
|
| ax.set_title("Hyperparam Sensitivity", fontsize=12, fontweight='bold')
|
| for (i, j), z in np.ndenumerate(data):
|
| ax.text(j, i, f'{z}%', ha='center', va='center')
|
|
|
| def plot_action_persistence(ax):
|
| """Dynamics: Action Persistence (Frame Skipping)."""
|
| ax.axis('off')
|
| ax.set_title("Action Persistence (k=4)", fontsize=12, fontweight='bold')
|
| for i in range(2):
|
| ax.add_patch(plt.Rectangle((0.1, 0.6-i*0.4), 0.8, 0.2, fill=False))
|
| ax.text(0.5, 0.7-i*0.4, f"Action A_{i}", ha='center')
|
| for j in range(4):
|
| ax.add_patch(plt.Rectangle((0.1+j*0.2, 0.6-i*0.4), 0.2, 0.2, fill=True, alpha=0.2))
|
| ax.text(0.5, 0.45, "Repeat Action for k frames", ha='center', color='blue', fontsize=8)
|
|
|
| def plot_muzero_search_tree(ax):
|
| """Model-Based: MuZero Search Tree with dynamics."""
|
| ax.axis('off')
|
| ax.set_title("MuZero Search Tree", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "Node $s$", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.3, 0.5, "Dyn $g$", bbox=dict(fc="lavender"), ha='center')
|
| ax.text(0.3, 0.1, "Pred $f$", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.3, 0.6), xytext=(0.5, 0.85), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.3, 0.2), xytext=(0.3, 0.4), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_policy_distillation(ax):
|
| """Deep RL: Policy Distillation (Teacher-Student)."""
|
| ax.axis('off')
|
| ax.set_title("Policy Distillation", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"Teacher $\pi_T$", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.8, 0.5, r"Student $\pi_S$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("KL-Divergence Loss", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->", lw=2, color='red'))
|
|
|
| def plot_decision_transformer_tokens(ax):
|
| """Transformers: Token Sequence (DT/TT)."""
|
| ax.axis('off')
|
| ax.set_title("Decision Transformer Tokens", fontsize=12, fontweight='bold')
|
| tokens = [r"$\hat{R}_t$", "$s_t$", "$a_t$", r"$\hat{R}_{t+1}$", "$s_{t+1}$"]
|
| for i, t in enumerate(tokens):
|
| ax.text(0.1+i*0.2, 0.5, t, bbox=dict(boxstyle="round", fc="white"))
|
| ax.annotate("causal attention", xy=(0.5, 0.7), xytext=(0.5, 0.6), annotation_clip=False)
|
|
|
| def plot_performance_profiles_rliable(ax):
|
| """Evaluation: Success Probability Profiles (rliable)."""
|
| x = np.linspace(0, 1, 100)
|
| y1 = x**2
|
| y2 = np.sqrt(x)
|
| ax.plot(x, y1, label="Algo A")
|
| ax.plot(x, y2, label="Algo B")
|
| ax.set_title("Performance Profiles", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Normalized Score")
|
| ax.set_ylabel("Probability of higher score")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_safety_shielding(ax):
|
| """Safety RL: Action Shielding / Constraints."""
|
| ax.axis('off')
|
| ax.set_title("Safety Shielding", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.4, fill=True, color='red', alpha=0.1))
|
| ax.text(0.5, 0.5, "Forbidden\nRegion", ha='center', color='red')
|
| ax.annotate("Shielded Action", xy=(0.2, 0.2), xytext=(0.4, 0.4), arrowprops=dict(arrowstyle="->", color='green', lw=2))
|
|
|
| def plot_automated_curriculum(ax):
|
| """Training: Automated Curriculum Difficulty."""
|
| t = np.arange(100)
|
| difficulty = 1.0 / (1.0 + np.exp(-0.05 * (t - 50)))
|
| performance = 0.8 / (1.0 + np.exp(-0.05 * (t - 40)))
|
| ax.plot(t, difficulty, label="Task Difficulty", color='black')
|
| ax.plot(t, performance, '--', label="Agent Performance", color='blue')
|
| ax.set_title("Automated Curriculum", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_domain_randomization(ax):
|
| """Sim-to-Real: Domain Randomization parameter distribution."""
|
| params = np.random.normal(1.0, 0.3, 1000)
|
| ax.hist(params, bins=30, color='orange', alpha=0.6)
|
| ax.set_title("Domain Randomization ($P(\\mu)$)", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Friction / Mass Parameter")
|
|
|
| def plot_rlhf_flow(ax):
|
| """Alignment: RL with Human Feedback (RLHF)."""
|
| ax.axis('off')
|
| ax.set_title("RLHF Flow Diagram", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.8, "Human Pref", bbox=dict(fc="salmon"), ha='center')
|
| ax.text(0.5, 0.8, "Reward Model", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.9, 0.8, "Fine-tuned Policy", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.8), xytext=(0.2, 0.8), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.8), xytext=(0.6, 0.8), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("PPO Update", xy=(0.5, 0.5), xytext=(0.9, 0.7), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=0.3"))
|
|
|
| def plot_successor_representations(ax):
|
| """Neuro-inspired RL: Successor Representation (SR) Matrix M."""
|
| M = np.zeros((10, 10))
|
| for i in range(10):
|
| for j in range(10):
|
| M[i, j] = 0.9**abs(i-j)
|
| ax.imshow(M, cmap='viridis')
|
| ax.set_title("Successor Representation $M$", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("State $j$")
|
| ax.set_ylabel("State $i$")
|
|
|
| def plot_maxent_irl_trajectories(ax):
|
| """IRL: MaxEnt IRL (Log-probability of trajectories)."""
|
| ax.axis('off')
|
| ax.set_title("MaxEnt IRL Distribution", fontsize=12, fontweight='bold')
|
| for i in range(5):
|
| alpha = 0.1 + i*0.2
|
| ax.plot([0, 1], [0.5, 0.5+0.1*i], color='blue', alpha=alpha)
|
| ax.plot([0, 1], [0.5, 0.5-0.1*i], color='blue', alpha=alpha)
|
| ax.text(0.5, 0.8, r"$P(\tau) \propto \exp(R(\tau))$", ha='center', fontsize=12)
|
|
|
| def plot_information_bottleneck(ax):
|
| """Theory: Information Bottleneck in RL."""
|
| ax.axis('off')
|
| ax.set_title("Information Bottleneck", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "S", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.5, 0.5, "Z", bbox=dict(boxstyle="circle", fc="gold"), ha='center')
|
| ax.text(0.9, 0.5, "A", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.annotate("Compress", xy=(0.4, 0.5), xytext=(0.15, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Extract", xy=(0.85, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.2, r"$\min I(S;Z)$ s.t. $I(Z;A) \geq I_c$", ha='center', fontsize=8)
|
|
|
| def plot_es_population_distribution(ax):
|
| """Evolutionary Strategies: ES Population Distribution."""
|
| np.random.seed(0)
|
| mu = [0, 0]
|
| points = np.random.randn(50, 2) * 0.5 + mu
|
| ax.scatter(points[:, 0], points[:, 1], color='blue', alpha=0.4, label="Population")
|
| ax.scatter(mu[0], mu[1], color='red', marker='x', label=r"$\mu$")
|
| ax.annotate("Gradient Estimate", xy=(1.0, 1.0), xytext=(0, 0), arrowprops=dict(arrowstyle="->", color='red'))
|
| ax.set_title("ES Population Update", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_cbf_safe_set(ax):
|
| """Safety RL: Control Barrier Function (CBF) Safe Set."""
|
| ax.axis('off')
|
| ax.set_title("CBF Safe Set Boundary", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.35, fill=False, color='black', lw=2))
|
| ax.text(0.5, 0.5, r"Safe Set $h(s) \geq 0$", ha='center')
|
| ax.text(0.5, 0.1, "Unsafe $h(s) < 0$", ha='center', color='red')
|
| ax.annotate("", xy=(0.8, 0.8), xytext=(0.6, 0.6), arrowprops=dict(arrowstyle="->", color='blue'))
|
| ax.text(0.75, 0.65, r"$\nabla h$", color='blue')
|
|
|
| def plot_count_based_exploration(ax):
|
| """Exploration: Count-based Heatmap N(s)."""
|
| grid = np.random.poisson(2, (10, 10))
|
| grid[0, 0] = 50; grid[9, 9] = 1
|
| im = ax.imshow(grid, cmap='hot')
|
| ax.set_title("Visit Counts $N(s)$", fontsize=12, fontweight='bold')
|
| plt.colorbar(im, ax=ax, label="Visits")
|
|
|
| def plot_thompson_sampling(ax):
|
| """Exploration: Thompson Sampling Posterior Distribution."""
|
| x = np.linspace(0, 1, 100)
|
| import scipy.stats as stats
|
| y1 = stats.beta.pdf(x, 2, 5)
|
| y2 = stats.beta.pdf(x, 10, 4)
|
| ax.plot(x, y1, label="Action 1 (Uncertain)")
|
| ax.plot(x, y2, label="Action 2 (Certain)")
|
| ax.fill_between(x, y1, alpha=0.2)
|
| ax.fill_between(x, y2, alpha=0.2)
|
| ax.set_title("Thompson Sampling Posteriors", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_adversarial_rl_interaction(ax):
|
| """Multi-Agent: Adversarial RL (Protaganist vs Antagonist)."""
|
| ax.axis('off')
|
| ax.set_title("Adversarial RL Interaction", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Protaganist", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, "Antagonist", bbox=dict(fc="salmon"), ha='center')
|
| ax.annotate("Force Distortion", xy=(0.35, 0.5), xytext=(0.65, 0.5), arrowprops=dict(arrowstyle="->", color='red'))
|
| ax.annotate("Policy Update", xy=(0.5, 0.8), xytext=(0.5, 0.6), arrowprops=dict(arrowstyle="<-", connectionstyle="arc3,rad=-0.3"))
|
|
|
| def plot_hierarchical_subgoals(ax):
|
| """Hierarchical RL: Subgoal Trajectory Waypoints."""
|
| ax.set_title("Subgoal Trajectory", fontsize=12, fontweight='bold')
|
| ax.plot([0, 1], [0, 1], 'k--', alpha=0.3)
|
| ax.scatter([0, 0.3, 0.7, 1], [0, 0.4, 0.6, 1], c=['black', 'red', 'red', 'gold'], s=100)
|
| ax.text(0.3, 0.45, "Subgoal 1", color='red', fontsize=8)
|
| ax.text(0.7, 0.65, "Subgoal 2", color='red', fontsize=8)
|
| ax.text(1, 1.1, "Final Goal", color='gold', fontweight='bold', ha='center')
|
|
|
| def plot_offline_distribution_shift(ax):
|
| """Offline RL: Distribution Shift (Shift between D and pi)."""
|
| x = np.linspace(-5, 5, 200)
|
| d = np.exp(-(x+1)**2 / 2)
|
| pi = np.exp(-(x-2)**2 / 1.5)
|
| ax.plot(x, d, label=r"Offline Dataset $\mathcal{D}$", color='grey')
|
| ax.plot(x, pi, label=r"Learned Policy $\pi$", color='blue')
|
| ax.fill_between(x, 0, d, color='grey', alpha=0.1)
|
| ax.fill_between(x, 0, pi, color='blue', alpha=0.1)
|
| ax.set_title("Action Distribution Shift", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_rnd_curiosity(ax):
|
| """Exploration: Random Network Distillation (RND)."""
|
| ax.axis('off')
|
| ax.set_title("RND: Predictor vs Target", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "State $s$", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.3, 0.5, "Fixed Target Net", bbox=dict(fc="lightgrey"), ha='center', fontsize=8)
|
| ax.text(0.7, 0.5, "Predictor Net", bbox=dict(fc="ivory"), ha='center', fontsize=8)
|
| ax.text(0.5, 0.2, "MSE Error = Intrinsic Reward", ha='center', color='red', fontsize=9)
|
| ax.annotate("", xy=(0.3, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_bcq_offline_constraint(ax):
|
| """Offline RL: Batch-Constrained Q-learning (BCQ)."""
|
| ax.axis('off')
|
| ax.set_title("BCQ: Action Constraint", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.35, fill=True, color='blue', alpha=0.1))
|
| ax.text(0.5, 0.5, "Dataset Action\nDistribution", ha='center', color='blue')
|
| ax.annotate("Constrained Action", xy=(0.4, 0.45), xytext=(0.2, 0.2), arrowprops=dict(arrowstyle="->", lw=2))
|
| ax.text(0.5, 0.1, r"$\max Q(s, a)$ s.t. $a \in \mathcal{D}$", ha='center', fontsize=9)
|
|
|
| def plot_pbt_evolution(ax):
|
| """Training: Population-Based Training (PBT)."""
|
| ax.axis('off')
|
| ax.set_title("Population-Based Training", fontsize=12, fontweight='bold')
|
| for i in range(3):
|
| ax.plot([0.1, 0.9], [0.8-i*0.3, 0.8-i*0.3], 'grey', alpha=0.3)
|
| ax.text(0.1, 0.8-i*0.3, f"Agent {i+1}", ha='right')
|
| ax.scatter([0.2, 0.5, 0.8], [0.8-i*0.3, 0.8-i*0.3, 0.8-i*0.3], color='blue')
|
| ax.annotate("Exploit & Perturb", xy=(0.5, 0.2), xytext=(0.5, 0.5), arrowprops=dict(arrowstyle="->", color='red'))
|
|
|
| def plot_recurrent_state_flow(ax):
|
| """Deep RL: Recurrent State Flow (DRQN/R2D2)."""
|
| ax.axis('off')
|
| ax.set_title("Recurrent $h_t$ Flow", fontsize=12, fontweight='bold')
|
| for i in range(3):
|
| ax.text(0.2+i*0.3, 0.5, f"Cell {i}", bbox=dict(fc="ivory"), ha='center')
|
| if i < 2:
|
| ax.annotate("", xy=(0.35+i*0.3, 0.5), xytext=(0.25+i*0.3, 0.5), arrowprops=dict(arrowstyle="->", color='blue'))
|
| ax.text(0.3+i*0.3, 0.55, rf"$h_{i}$", color='blue', fontsize=8)
|
|
|
| def plot_belief_state_pomdp(ax):
|
| """Theory: Belief State in POMDPs."""
|
| x = np.linspace(0, 1, 100)
|
| y = np.exp(-(x-0.3)**2 / 0.02) + 0.3*np.exp(-(x-0.8)**2 / 0.01)
|
| ax.plot(x, y, color='purple')
|
| ax.fill_between(x, y, alpha=0.2, color='purple')
|
| ax.set_title(r"Belief State $b(s)$", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("State Space")
|
| ax.set_ylabel("Probability")
|
|
|
| def plot_pareto_front_morl(ax):
|
| """Multi-Objective RL: Pareto Front."""
|
| np.random.seed(42)
|
| x = np.random.rand(50)
|
| y = np.random.rand(50)
|
| ax.scatter(x, y, alpha=0.3, color='grey')
|
|
|
| px = np.sort(x)[-10:]
|
| py = np.sort(y)[-10:][::-1]
|
| ax.plot(px, py, 'r-o', label="Pareto Front")
|
| ax.set_title("Multi-Objective Pareto Front", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Reward A")
|
| ax.set_ylabel("Reward B")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_differential_value_average_reward(ax):
|
| """Theory: Differential Value (Average Reward RL)."""
|
| t = np.arange(100)
|
| v = np.sin(0.2*t) + 0.05*t
|
| rho = 0.05
|
| ax.plot(t, v, label="Value $V(s_t)$")
|
| ax.plot(t, rho*t, '--', label=r"Gain $\rho \cdot t$", color='red')
|
| ax.set_title("Differential Value $v(s)$", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_distributed_rl_cluster(ax):
|
| """Infrastructure: Distributed RL Cluster (Ray/RLLib)."""
|
| ax.axis('off')
|
| ax.set_title("Distributed RL Cluster", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Learner / GPU", bbox=dict(boxstyle="round", fc="gold"), ha='center')
|
| ax.text(0.5, 0.5, "Replay Buffer", bbox=dict(fc="lightgrey"), ha='center')
|
| for i in range(3):
|
| ax.text(0.2+i*0.3, 0.2, f"Worker {i+1}", bbox=dict(fc="ivory"), ha='center', fontsize=8)
|
| ax.annotate("", xy=(0.5, 0.45), xytext=(0.2+i*0.3, 0.25), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.75), xytext=(0.5, 0.55), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_neuroevolution_topology(ax):
|
| """Evolutionary RL: Topology Evolution (NEAT)."""
|
| ax.axis('off')
|
| ax.set_title("Neuroevolution Topology", fontsize=12, fontweight='bold')
|
| nodes = [(0.2, 0.5), (0.5, 0.8), (0.5, 0.2), (0.8, 0.5)]
|
| for p in nodes: ax.text(p[0], p[1], "", bbox=dict(boxstyle="circle", fc="white"))
|
|
|
| ax.annotate("", xy=nodes[1], xytext=nodes[0], arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=nodes[2], xytext=nodes[0], arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=nodes[3], xytext=nodes[1], arrowprops=dict(arrowstyle="->"))
|
|
|
| ax.text(0.5, 0.5, "New Node", bbox=dict(boxstyle="circle", fc="yellow"), ha='center', fontsize=7)
|
| ax.annotate("", xy=(0.5, 0.5), xytext=nodes[0], arrowprops=dict(arrowstyle="->", color='red', ls='--'))
|
|
|
| def plot_ewc_elastic_weights(ax):
|
| """Continual RL: Elastic Weight Consolidation (EWC)."""
|
| ax.axis('off')
|
| ax.set_title("EWC Elastic Constraint", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.3, 0.5), 0.2, color='blue', alpha=0.2, label="Task A"))
|
| ax.add_patch(plt.Circle((0.7, 0.5), 0.2, color='red', alpha=0.2, label="Task B"))
|
| ax.annotate("", xy=(0.5, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=0.3"))
|
| ax.text(0.5, 0.7, "Spring Constraint", color='darkgreen', ha='center', fontsize=9)
|
|
|
| def plot_successor_features(ax):
|
| """Theory: Successor Features (SF)."""
|
| ax.axis('off')
|
| ax.set_title(r"Successor Features $\psi$", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"Features $\phi(s)$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.8, 0.5, r"SF $\psi(s)$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate(r"$\sum \gamma^t \phi(s_t)$", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->", lw=2))
|
|
|
| def plot_adversarial_state_noise(ax):
|
| r"""Safety: Adversarial State Noise ($s + \delta$)."""
|
| ax.axis('off')
|
| ax.set_title("Adversarial Perturbation", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "State $s$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.5, "+", fontsize=20, ha='center')
|
| ax.text(0.8, 0.5, r"Noise $\delta$", bbox=dict(fc="salmon"), ha='center')
|
| ax.annotate("Target: Wrong Action!", xy=(0.5, 0.2), xytext=(0.5, 0.4), arrowprops=dict(arrowstyle="->", color='red'))
|
|
|
| def plot_behavioral_cloning_il(ax):
|
| """Imitation: Behavioral Cloning (BC)."""
|
| ax.axis('off')
|
| ax.set_title("Behavioral Cloning Flow", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Expert Data\n$(s^*, a^*)$", bbox=dict(fc="gold"), ha='center', fontsize=8)
|
| ax.text(0.5, 0.5, "Supervised\nLearning", bbox=dict(fc="ivory"), ha='center', fontsize=8)
|
| ax.text(0.9, 0.5, r"Clone Policy\n$\pi_{BC}$", bbox=dict(fc="lightgrey"), ha='center', fontsize=8)
|
| ax.annotate("", xy=(0.35, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_relational_graph_state(ax):
|
| """Relational RL: Graph-based State Representation."""
|
| ax.axis('off')
|
| ax.set_title("Relational Graph State", fontsize=12, fontweight='bold')
|
| pos = {1: (0.3, 0.7), 2: (0.7, 0.7), 3: (0.5, 0.3)}
|
| for k, p in pos.items():
|
| ax.text(p[0], p[1], f"Obj {k}", bbox=dict(boxstyle="round", fc="lightblue"), ha='center')
|
| edges = [(1, 2), (2, 3), (3, 1)]
|
| for u, v in edges:
|
| ax.annotate("relation", xy=pos[v], xytext=pos[u], arrowprops=dict(arrowstyle="-", color='grey', ls=':'), ha='center')
|
|
|
| def plot_quantum_rl_circuit(ax):
|
| """Quantum RL: Parameterized Quantum Circuit (PQC) Policy."""
|
| ax.axis('off')
|
| ax.set_title("Quantum Policy (PQC)", fontsize=12, fontweight='bold')
|
| ax.plot([0.1, 0.9], [0.7, 0.7], 'k', lw=1)
|
| ax.plot([0.1, 0.9], [0.3, 0.3], 'k', lw=1)
|
| ax.text(0.2, 0.7, r"$|0\rangle$", ha='right')
|
| ax.text(0.2, 0.3, r"$|0\rangle$", ha='right')
|
|
|
| ax.text(0.4, 0.7, r"$R_y(\theta)$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.6, 0.5, "CNOT", bbox=dict(fc="gold"), ha='center')
|
| ax.plot([0.6, 0.6], [0.3, 0.7], 'k-o')
|
| ax.text(0.8, 0.7, r"$\mathcal{M}$", bbox=dict(boxstyle="square", fc="lightgrey"), ha='center')
|
|
|
| def plot_symbolic_expression_tree(ax):
|
| """Symbolic RL: Policy as a Mathematical Expression Tree."""
|
| ax.axis('off')
|
| ax.set_title("Symbolic Policy Tree", fontsize=12, fontweight='bold')
|
| nodes = {0:(0.5, 0.8, "+"), 1:(0.3, 0.5, "*"), 2:(0.7, 0.5, "exp"), 3:(0.2, 0.2, "s"), 4:(0.4, 0.2, "2.5"), 5:(0.7, 0.2, "s")}
|
| edges = [(0,1), (0,2), (1,3), (1,4), (2,5)]
|
| for k, (x, y, t) in nodes.items():
|
| ax.text(x, y, t, bbox=dict(boxstyle="circle", fc="ivory"), ha='center')
|
| for u, v in edges:
|
| ax.annotate("", xy=nodes[v][:2], xytext=nodes[u][:2], arrowprops=dict(arrowstyle="-"))
|
|
|
| def plot_differentiable_physics_gradient(ax):
|
| """Control: Differentiable Physics Gradient Flow."""
|
| ax.axis('off')
|
| ax.set_title("Diff-Physics Gradient", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Policy", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, "Diff-Sim\nDynamics", bbox=dict(fc="gold", boxstyle="round"), ha='center')
|
| ax.text(0.9, 0.5, "Loss", bbox=dict(fc="salmon"), ha='center')
|
|
|
| ax.annotate("", xy=(0.35, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.5), xytext=(0.65, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| ax.annotate("$\nabla$ gradient", xy=(0.15, 0.4), xytext=(0.85, 0.4), arrowprops=dict(arrowstyle="->", color='red', connectionstyle="arc3,rad=-0.2"))
|
|
|
| def plot_marl_communication_channel(ax):
|
| """MARL: Communication Channel (CommNet/DIAL)."""
|
| ax.axis('off')
|
| ax.set_title("Multi-Agent Comm Channel", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.8, "Agent A", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.8, "Agent B", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.2, "Task Goal", bbox=dict(fc="lightgrey"), ha='center')
|
|
|
| ax.annotate("Message $m_{A \to B}$", xy=(0.7, 0.8), xytext=(0.3, 0.8), arrowprops=dict(arrowstyle="->", ls="--", color='purple'))
|
| ax.annotate("", xy=(0.2, 0.45), xytext=(0.2, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.45), xytext=(0.8, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_lagrangian_multiplier_landscape(ax):
|
| """Safety: Lagrangian Constraint Optimization."""
|
| x = np.linspace(-2, 2, 100); y = np.linspace(-2, 2, 100)
|
| X, Y = np.meshgrid(x, y); Z = X**2 + Y**2
|
| ax.contour(X, Y, Z, levels=10, alpha=0.3)
|
| ax.axvline(x=0.5, color='red', ls='--', label=r"Constraint $g(s) \leq 0$")
|
| ax.scatter([1.0], [1.0], color='blue', label="Unconstrained Min")
|
| ax.scatter([0.5], [0.0], color='green', label="Constrained Min")
|
| ax.set_title("Lagrangian Constrained Opt", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=7, loc='upper left')
|
|
|
| def plot_maxq_task_hierarchy(ax):
|
| """HRL: MAXQ Recursive Task Decomposition."""
|
| ax.axis('off')
|
| ax.set_title("MAXQ Task Hierarchy", fontsize=12, fontweight='bold')
|
|
|
| ax.text(0.5, 0.9, "Root Task", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.3, 0.6, "GetFuel", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.7, 0.6, "DeliverCargo", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.3, 0.3, "Navigate", bbox=dict(fc="lightgrey"), ha='center', fontsize=8)
|
| ax.text(0.7, 0.3, "Unload", bbox=dict(fc="lightgrey"), ha='center', fontsize=8)
|
|
|
| ax.annotate("", xy=(0.3, 0.65), xytext=(0.45, 0.85), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.65), xytext=(0.55, 0.85), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.3, 0.35), xytext=(0.3, 0.55), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.35), xytext=(0.7, 0.55), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_react_cycle_thinking(ax):
|
| """Agentic LLM: ReAct Loop (Thought-Action-Observation)."""
|
| ax.axis('off')
|
| ax.set_title(r"ReAct Cycle: $T \to A \to O$", fontsize=12, fontweight='bold')
|
| steps = ["Thought", "Action", "Observation"]
|
| colors = ["ivory", "lightblue", "lightgreen"]
|
| for i, s in enumerate(steps):
|
| angle = 2 * np.pi * i / 3
|
| x, y = 0.5 + 0.3*np.cos(angle), 0.5 + 0.3*np.sin(angle)
|
| ax.text(x, y, s, bbox=dict(boxstyle="round", fc=colors[i]), ha='center')
|
|
|
| ax.annotate("", xy=(0.2, 0.5), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=0.3"))
|
| ax.annotate("", xy=(0.5, 0.2), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=0.3"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.5, 0.2), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=0.3"))
|
|
|
| def plot_synaptic_plasticity_rl(ax):
|
| """Bio-inspired: Synaptic Plasticity (Hebbian RL/STDP)."""
|
| ax.axis('off')
|
| ax.set_title("Synaptic Plasticity RL", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.5, "Pre-neuron", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.7, 0.5, "Post-neuron", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.plot([0.35, 0.65], [0.5, 0.5], 'k', lw=4, label="Synapse $w$")
|
| ax.text(0.5, 0.6, r"$\Delta w \propto \delta \cdot x_{pre} \cdot x_{post}$", color='red', ha='center', fontsize=10)
|
| ax.annotate(r"TD Error $\delta$", xy=(0.5, 0.5), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="->", color='red'))
|
|
|
| def plot_guided_policy_search_gps(ax):
|
| """Control: Guided Policy Search (GPS)."""
|
| ax.axis('off')
|
| ax.set_title("Guided Policy Search (GPS)", fontsize=12, fontweight='bold')
|
| ax.plot([0.1, 0.9], [0.7, 0.8], 'b', label=r"Optimal Trajectory $\tau^*$")
|
| ax.plot([0.1, 0.9], [0.6, 0.6], 'r--', label=r"Current Policy $\pi_\theta$")
|
| ax.annotate("Minimize KL", xy=(0.5, 0.6), xytext=(0.5, 0.72), arrowprops=dict(arrowstyle="<->"))
|
| ax.legend(fontsize=8, loc='lower right')
|
|
|
| def plot_sim2real_jitter_latency(ax):
|
| """Robotics: Sim-to-Real Jitter & Latency Analysis."""
|
| t = np.linspace(0, 10, 100)
|
| ideal = np.sin(t)
|
| jitter = ideal + 0.2*np.random.randn(100)
|
| ax.plot(t, ideal, 'g', alpha=0.5, label="Simulator (Ideal)")
|
| ax.step(t + 0.3, jitter, 'r', label="Real Robot (Latency+Jitter)")
|
| ax.set_title("Sim-to-Real Temporal Mismatch", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Time (s)")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_ddpg_deterministic_gradient(ax):
|
| """Deterministic Policy Gradient (DDPG)."""
|
| ax.axis('off')
|
| ax.set_title("DDPG Gradient Flow", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"$\pi_\theta(s)$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, r"$Q_w(s, a)$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate(r"$\nabla_\theta J \approx \nabla_a Q(s,a)|_{a=\pi(s)} \nabla_\theta \pi_\theta(s)$", xy=(0.5, 0.2), xytext=(0.5, 0.4), arrowprops=dict(arrowstyle="->", color='red'), ha='center', fontsize=9)
|
| ax.annotate("action", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_dreamer_latent_rollout(ax):
|
| """Model-Based RL: Dreamer Latent imagination."""
|
| ax.axis('off')
|
| ax.set_title("Dreamer Latent imagination", fontsize=12, fontweight='bold')
|
| for i in range(3):
|
| ax.text(0.2 + i*0.3, 0.5, f"$z_{i}$", bbox=dict(boxstyle="circle", fc="lightgreen"), ha='center')
|
| if i < 2:
|
| ax.annotate("", xy=(0.35 + i*0.3, 0.5), xytext=(0.25 + i*0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.3 + i*0.3, 0.7, r"$\hat{a}$", ha='center')
|
| ax.text(0.5, 0.2, r"Policy $\pi(z)$ learned in latent space", fontsize=9, ha='center')
|
|
|
| def plot_unreal_auxiliary_tasks(ax):
|
| """Deep RL: UNREAL Architecture (Auxiliary Tasks)."""
|
| ax.axis('off')
|
| ax.set_title("UNREAL Auxiliary Tasks", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Base Agent (A3C)", bbox=dict(fc="ivory"), ha='center')
|
| tasks = ["Pixel Control", "Value Replay", "Reward Prediction"]
|
| for i, t in enumerate(tasks):
|
| ax.text(0.2 + i*0.3, 0.4, t, bbox=dict(fc="orange", alpha=0.3), ha='center', fontsize=8)
|
| ax.annotate("", xy=(0.2+i*0.3, 0.5), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->", ls=':'))
|
| ax.text(0.5, 0.1, "Shared Representation Learning", fontweight='bold', ha='center', fontsize=9)
|
|
|
| def plot_iql_expectile_loss(ax):
|
| """Offline RL: Implicit Q-Learning (IQL) Expectile."""
|
| x = np.linspace(-2, 2, 100)
|
| tau = 0.8
|
| loss = np.where(x > 0, tau * x**2, (1-tau) * x**2)
|
| ax.plot(x, loss, color='purple', lw=2)
|
| ax.set_title(r"IQL Expectile Loss $L_\tau$", fontsize=12, fontweight='bold')
|
| ax.axvline(0, color='black', alpha=0.3)
|
| ax.text(1, 1, r"$\tau=0.8$", color='purple')
|
|
|
| def plot_prioritized_sweeping(ax):
|
| """Model-Based: Prioritized Sweeping."""
|
| ax.axis('off')
|
| ax.set_title("Prioritized Sweeping", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.8, "State $s$", bbox=dict(fc="white"), ha='center')
|
| ax.text(0.8, 0.2, "Priority Queue", bbox=dict(boxstyle="sawtooth", fc="gold"), ha='center')
|
| ax.annotate(r"TD Error $|\delta|$", xy=(0.7, 0.3), xytext=(0.3, 0.7), arrowprops=dict(arrowstyle="->", color='red'))
|
| ax.text(0.5, 0.5, "Update most affected states first", rotation=-35, fontsize=8)
|
|
|
| def plot_dagger_expert_loop(ax):
|
| """Imitation: DAgger (Dataset Aggregation)."""
|
| ax.axis('off')
|
| ax.set_title("DAgger Expert Loop", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.7, r"Learner $\pi_\theta$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.7, r"Expert $\pi^*$", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.3, r"Dataset $\mathcal{D}$", bbox=dict(boxstyle="round", fc="ivory"), ha='center')
|
| ax.annotate("Collect", xy=(0.5, 0.4), xytext=(0.2, 0.6), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Relabel", xy=(0.8, 0.6), xytext=(0.5, 0.4), arrowprops=dict(arrowstyle="<-"))
|
| ax.annotate("Train", xy=(0.25, 0.65), xytext=(0.4, 0.35), arrowprops=dict(arrowstyle="->", color='blue'))
|
|
|
| def plot_spr_self_prediction(ax):
|
| """Deep RL: Self-Predictive Representations (SPR)."""
|
| ax.axis('off')
|
| ax.set_title("SPR: Self-Prediction", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Encoder", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.8, 0.7, "Target Latent", bbox=dict(fc="gold", alpha=0.3), ha='center')
|
| ax.text(0.8, 0.3, "Predicted Latent", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate("", xy=(0.7, 0.7), xytext=(0.3, 0.55), arrowprops=dict(arrowstyle="->", ls='--'))
|
| ax.annotate("", xy=(0.7, 0.3), xytext=(0.3, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.9, 0.5, "Consistency Loss", rotation=90, color='red', fontsize=8)
|
|
|
| def plot_joint_action_space(ax):
|
| """MARL: Joint Action Space $A_1 \times A_2$."""
|
| ax.set_title(r"Joint Action Space $A_1 \times A_2$", fontsize=12, fontweight='bold')
|
| for x in range(3):
|
| for y in range(3):
|
| ax.scatter(x, y, color='blue', alpha=0.5)
|
| ax.text(x, y+0.1, f"($a^k_{x}, a^j_{y}$)", fontsize=7, ha='center')
|
| ax.set_xlabel("Agent 1 Actions")
|
| ax.set_ylabel("Agent 2 Actions")
|
| ax.set_xticks([0,1,2]); ax.set_yticks([0,1,2])
|
|
|
| def plot_dec_pomdp_graph(ax):
|
| """MARL: Dec-POMDP Formal Model."""
|
| ax.axis('off')
|
| ax.set_title("Dec-POMDP Model", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Global State $s$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.2, 0.4, "Obs $o_1$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.4, "Obs $o_2$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.1, "Joint Reward $r$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate("", xy=(0.2, 0.5), xytext=(0.45, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.55, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.45, 0.15), xytext=(0.2, 0.35), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.55, 0.15), xytext=(0.8, 0.35), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_bisimulation_metric(ax):
|
| """Theory: State Bisimulation Metric."""
|
| ax.axis('off')
|
| ax.set_title("Bisimulation Metric", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.6, "$s_1$", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.text(0.7, 0.6, "$s_2$", bbox=dict(boxstyle="circle", fc="white"), ha='center')
|
| ax.annotate("$d(s_1, s_2)$", xy=(0.65, 0.6), xytext=(0.35, 0.6), arrowprops=dict(arrowstyle="<->", color='purple'))
|
| ax.text(0.5, 0.2, "States are equivalent if rewards and\ntransitions to equivalent states match", ha='center', fontsize=8)
|
|
|
| def plot_reward_shaping_phi(ax):
|
| """Theory: Potential-Based Reward Shaping."""
|
| ax.axis('off')
|
| ax.set_title("Potential-Based Reward Shaping", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "$s$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.8, 0.5, "$s'$", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.7, r"$\gamma \Phi(s') - \Phi(s)$", color='blue', ha='center')
|
| ax.text(0.5, 0.3, "Added to environmental reward $r$", fontsize=8, ha='center')
|
|
|
| def plot_transfer_rl_source_target(ax):
|
| """Training: Transfer RL (Source to Target)."""
|
| ax.axis('off')
|
| ax.set_title("Transfer RL: Source to Target", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.7, r"Source Task $\mathcal{T}_A$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.7, 0.3, r"Target Task $\mathcal{T}_B$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("Knowledge Transfer\n(Weights/Expert Data)", xy=(0.6, 0.4), xytext=(0.4, 0.6), arrowprops=dict(arrowstyle="->", lw=2, color='orange'), ha='center')
|
|
|
| def plot_multi_task_backbone(ax):
|
| """Deep RL: Multi-Task Architecture."""
|
| ax.axis('off')
|
| ax.set_title("Multi-Task Backbone Arch", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "State Input", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.5, 0.5, "Shared Backbone", bbox=dict(fc="cornflowerblue"), ha='center')
|
| ax.text(0.2, 0.2, "Task 1 Head", bbox=dict(fc="orange", alpha=0.5), ha='center')
|
| ax.text(0.8, 0.2, "Task N Head", bbox=dict(fc="orange", alpha=0.5), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="<-"))
|
| ax.annotate("", xy=(0.25, 0.3), xytext=(0.45, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.3), xytext=(0.55, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_contextual_bandit_pipeline(ax):
|
| """Bandits: Contextual Bandit Pipeline."""
|
| ax.axis('off')
|
| ax.set_title("Contextual Bandit Pipeline", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, r"Context $x$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, r"Policy $\pi(a|x)$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.9, 0.5, r"Reward $r$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_regret_bounds_theoretical(ax):
|
| """Theory: Regret Upper/Lower Bounds."""
|
| t = np.linspace(1, 100, 100)
|
| ax.plot(t, np.sqrt(t), label=r"Upper Bound $O(\sqrt{T})$", color='red')
|
| ax.plot(t, np.log(t), label=r"Optimal Regret $O(\log T)$", color='blue')
|
| ax.set_title("Theoretical Regret Bounds", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Time $T$")
|
| ax.set_ylabel("Cumulative Regret")
|
| ax.legend()
|
|
|
| def plot_soft_q_heatmap(ax):
|
| """Value-based: Soft Q-Learning Heatmap."""
|
| data = np.random.randn(10, 10)
|
| soft_q = np.exp(data) / np.sum(np.exp(data))
|
| im = ax.imshow(soft_q, cmap='hot')
|
| plt.colorbar(im, ax=ax)
|
| ax.set_title("Soft Q Boltzmann Probabilities", fontsize=12, fontweight='bold')
|
|
|
| def plot_ad_rl_pipeline(ax):
|
| """Robotics: Autonomous Driving RL Pipeline."""
|
| ax.axis('off')
|
| ax.set_title("Autonomous Driving RL Pipeline", fontsize=12, fontweight='bold')
|
| modules = ["Sensors", "Perception (CNN)", "RL Policy", "Actuators"]
|
| for i, m in enumerate(modules):
|
| ax.text(0.25 + (i%2)*0.5, 0.7 - (i//2)*0.5, m, bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.7, 0.7), xytext=(0.3, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.35), xytext=(0.75, 0.6), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.3, 0.2), xytext=(0.7, 0.2), arrowprops=dict(arrowstyle="<-"))
|
|
|
| def plot_action_grad_comparison(ax):
|
| """Policy: Stochastic vs Deterministic Gradients."""
|
| ax.axis('off')
|
| ax.set_title("Action Gradient Types", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.7, r"Stochastic: $\nabla \log \pi(a|s) Q(s,a)$", color='blue', ha='center')
|
| ax.text(0.5, 0.3, r"Deterministic: $\nabla_a Q(s,a) \nabla \pi(s)$", color='red', ha='center')
|
| ax.text(0.5, 0.5, "vs", fontweight='bold', ha='center')
|
|
|
| def plot_irl_feature_matching(ax):
|
| """IRL: Feature Expectation Matching."""
|
| ax.axis('off')
|
| ax.set_title("IRL: Feature Expectation Matching", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"Expert $\mu(\pi^*)$", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.8, 0.5, r"Learner $\mu(\pi)$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate(r"$||\mu(\pi^*) - \mu(\pi)||_2 \leq \epsilon$", xy=(0.5, 0.2), ha='center', color='red')
|
| ax.annotate("", xy=(0.65, 0.5), xytext=(0.35, 0.5), arrowprops=dict(arrowstyle="<->", ls='--'))
|
|
|
| def plot_apprenticeship_learning_loop(ax):
|
| """Imitation: Apprenticeship Learning Loop."""
|
| ax.axis('off')
|
| ax.set_title("Apprenticeship Learning Loop", fontsize=12, fontweight='bold')
|
| nodes = ["Expert Demos", "Reward Learning", "Agent Policy", "Environment"]
|
| for i, n in enumerate(nodes):
|
| ax.text(0.5, 0.9 - i*0.25, n, bbox=dict(fc="ivory"), ha='center')
|
| if i < 3: ax.annotate("", xy=(0.5, 0.7 - i*0.25), xytext=(0.5, 0.8 - i*0.25), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("feedback", xy=(0.3, 0.9), xytext=(0.3, 0.15), arrowprops=dict(arrowstyle="->", connectionstyle="arc3,rad=-0.5"))
|
|
|
| def plot_active_inference_loop(ax):
|
| """Theoretical: Active Inference / Free Energy Loop."""
|
| ax.axis('off')
|
| ax.set_title("Active Inference Loop", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Internal Model (Generative)", bbox=dict(fc="cornflowerblue", alpha=0.3), ha='center')
|
| ax.text(0.5, 0.2, "External Environment", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("Action (Active Charge)", xy=(0.8, 0.25), xytext=(0.8, 0.75), arrowprops=dict(arrowstyle="<-", color='red'))
|
| ax.annotate("Perception (Surprise Min)", xy=(0.2, 0.75), xytext=(0.2, 0.25), arrowprops=dict(arrowstyle="<-", color='blue'))
|
| ax.text(0.5, 0.5, r"$\min F = D_{KL}(q||p)$", ha='center', fontweight='bold')
|
|
|
| def plot_bellman_residual_landscape(ax):
|
| """Theory: Bellman Residual Landscape."""
|
| X, Y = np.meshgrid(np.linspace(-2, 2, 20), np.linspace(-2, 2, 20))
|
| Z = (X**2 + Y**2) + 0.5 * np.sin(3*X)
|
| ax.contourf(X, Y, Z, cmap='magma')
|
| ax.set_title("Bellman Residual Landscape", fontsize=12, fontweight='bold')
|
|
|
| def plot_plan_to_explore_map(ax):
|
| """MBRL: Plan-to-Explore Uncertainty Map."""
|
| data = np.random.rand(10, 10)
|
| im = ax.imshow(data, cmap='YlOrRd')
|
| ax.set_title("Plan-to-Explore Uncertainty", fontsize=12, fontweight='bold')
|
| ax.text(2, 2, "Explored", color='black', fontsize=8)
|
| ax.text(7, 7, "Unknown", color='red', fontweight='bold', fontsize=8)
|
|
|
| def plot_robust_rl_uncertainty_set(ax):
|
| """Safety: Robust RL Uncertainty Set."""
|
| ax.axis('off')
|
| ax.set_title("Robust RL Uncertainty Set", fontsize=12, fontweight='bold')
|
| circle = plt.Circle((0.5, 0.5), 0.3, color='blue', alpha=0.1)
|
| ax.add_patch(circle)
|
| ax.text(0.5, 0.5, r"$\mathcal{P}$", fontsize=20, ha='center')
|
| ax.text(0.5, 0.1, r"$\min_\pi \max_{P \in \mathcal{P}} \mathbb{E}[R]$", ha='center', fontsize=12)
|
| ax.annotate("Nominal Model", xy=(0.5, 0.5), xytext=(0.2, 0.8), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_hpo_bayesian_opt_cycle(ax):
|
| """Training: HPO Bayesian Optimization Cycle."""
|
| ax.axis('off')
|
| ax.set_title("HPO Bayesian Opt Cycle", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Surrogate Model (GP)", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.2, "RL Objective Function", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("Select Hyperparams", xy=(0.7, 0.3), xytext=(0.7, 0.7), arrowprops=dict(arrowstyle="<-"))
|
| ax.annotate("Update Model", xy=(0.3, 0.7), xytext=(0.3, 0.3), arrowprops=dict(arrowstyle="<-"))
|
|
|
| def plot_slate_rl_reco_pipeline(ax):
|
| """Applied: Slate RL / Recommendation Pipeline."""
|
| ax.axis('off')
|
| ax.set_title("Slate RL Recommendation", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "User State", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.5, 0.5, "Slate Policy", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.9, 0.5, "Action (Items)", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.2, "Combinatorial Action Space", fontsize=8, ha='center')
|
|
|
| def plot_game_theory_fictitious_play(ax):
|
| """Multi-Agent: Fictitious Play Interaction."""
|
| ax.axis('off')
|
| ax.set_title("Fictitious Play Interaction", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.7, "Agent A (Best Response)", bbox=dict(fc="white"), ha='center')
|
| ax.text(0.8, 0.7, "Agent B (Best Response)", bbox=dict(fc="white"), ha='center')
|
| ax.text(0.5, 0.3, r"Empirical Frequency $\hat{\pi}$", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.45, 0.4), xytext=(0.25, 0.6), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.55, 0.4), xytext=(0.75, 0.6), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_universal_rl_framework(ax):
|
| """Conceptual: Universal RL Framework Diagram."""
|
| ax.axis('off')
|
| ax.set_title("Universal RL Framework", fontsize=12, fontweight='bold')
|
| rect = plt.Rectangle((0.15, 0.15), 0.7, 0.7, fill=False, ls='--')
|
| ax.add_patch(rect)
|
| ax.text(0.5, 0.5, "RL Agent\n(Algorithm + Model + Exp)", ha='center', fontweight='bold')
|
| ax.text(0.5, 0.9, "Problem Context", ha='center', color='grey')
|
| ax.text(0.5, 0.1, "Reward / Evaluation", ha='center', color='grey')
|
|
|
| def plot_offline_density_ratio(ax):
|
| """Offline RL: Density Ratio Estimation $w(s,a)$."""
|
| x = np.linspace(-3, 3, 100)
|
| pi_e = norm.pdf(x, 0, 1)
|
| pi_b = norm.pdf(x, 1, 1.5)
|
| ax.plot(x, pi_e, label=r"Policy $\pi_e$")
|
| ax.plot(x, pi_b, label=r"Behavior $\pi_b$", ls='--')
|
| ax.fill_between(x, pi_e / (pi_b + 1e-5), alpha=0.1, label="Ratio $w$")
|
| ax.set_title(r"Offline Density Ratio $w(s,a)$", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_continual_task_interference(ax):
|
| """Continual RL: Task Interference Heatmap."""
|
| data = np.eye(5) + 0.1 * np.random.randn(5, 5)
|
| data[1,0] = -0.5
|
| im = ax.imshow(data, cmap='coolwarm', vmin=-1, vmax=1)
|
| plt.colorbar(im, ax=ax)
|
| ax.set_title("Continual Task Interference", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Previously Learned Tasks"); ax.set_ylabel("Current Task")
|
|
|
| def plot_lyapunov_safe_set(ax):
|
| """Safety: Lyapunov Stability Set."""
|
| ax.set_title("Lyapunov Safe Set", fontsize=12, fontweight='bold')
|
| theta = np.linspace(0, 2*np.pi, 100)
|
| r = 1 + 0.2 * np.sin(4*theta)
|
| ax.fill(r * np.cos(theta), r * np.sin(theta), color='green', alpha=0.1, label="Invariant Set")
|
| ax.plot(r * np.cos(theta), r * np.sin(theta), color='green')
|
| ax.quiver(0.5, 0.5, -0.4, -0.4, color='red', scale=5, label="Energy Decrease")
|
| ax.legend(fontsize=8); ax.set_xlim(-1.5, 1.5); ax.set_ylim(-1.5, 1.5)
|
|
|
| def plot_molecular_rl_atoms(ax):
|
| """Applied: Molecular RL (Atoms)."""
|
| ax.set_title("Molecular RL (Atom State)", fontsize=12, fontweight='bold')
|
| for _ in range(5):
|
| pos = np.random.rand(2)
|
| circle = plt.Circle(pos, 0.05, color='blue', alpha=0.7)
|
| ax.add_patch(circle)
|
| ax.set_xlim(0, 1); ax.set_ylim(0, 1); ax.axis('off')
|
| ax.text(0.5, -0.05, "States = Atomic Coordinates", ha='center', fontsize=8)
|
|
|
| def plot_moe_multi_task_arch(ax):
|
| """Architecture: MoE for Multi-task."""
|
| ax.axis('off')
|
| ax.set_title("MoE Multi-task Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "Gating Network", bbox=dict(fc="orange"), ha='center')
|
| for i in range(3):
|
| ax.text(0.2 + i*0.3, 0.5, f"Expert {i+1}", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("", xy=(0.2 + i*0.3, 0.6), xytext=(0.5, 0.8), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.2, "Joint Output", bbox=dict(fc="lightgrey"), ha='center')
|
|
|
| def plot_cma_es_distribution(ax):
|
| """Direct Policy Search: CMA-ES Distribution."""
|
| x = np.random.randn(200, 2)
|
| ax.scatter(x[:,0], x[:,1], alpha=0.3, color='grey')
|
| circle = plt.Circle((0, 0), 1.5, fill=False, color='red', lw=2, label="Sample Ellipsoid")
|
| ax.add_patch(circle)
|
| ax.set_title("CMA-ES Policy Search", fontsize=12, fontweight='bold')
|
| ax.legend(fontsize=8)
|
|
|
| def plot_elo_rating_preference(ax):
|
| """Alignment: Elo Rating Preference Plot."""
|
| x = np.linspace(0, 10, 10)
|
| y = 1000 + 100 * np.log(x + 1) + 20 * np.random.randn(10)
|
| ax.step(x, y, color='purple', where='post')
|
| ax.set_title("Policy Elo Rating vs Experience", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Relative Training Time"); ax.set_ylabel("Elo Rating")
|
|
|
| def plot_shap_lime_attribution(ax):
|
| """Explainable RL: SHAP/LIME Attribution."""
|
| ax.set_title("Action Attribution (SHAP)", fontsize=12, fontweight='bold')
|
| feats = ["Dist to Goal", "Velocity", "Agent Pitch", "Sensor 4"]
|
| vals = [0.6, -0.3, 0.1, 0.05]
|
| colors = ['green' if v > 0 else 'red' for v in vals]
|
| ax.barh(feats, vals, color=colors)
|
| ax.set_xlabel("Contribution to Action probability")
|
|
|
| def plot_pearl_context_encoder(ax):
|
| """Meta-RL: Context Encoder (PEARL)."""
|
| ax.axis('off')
|
| ax.set_title("PEARL Context Encoder", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Experience batch\n(s, a, r, s')", bbox=dict(fc="ivory"), ha='center', fontsize=8)
|
| ax.text(0.5, 0.5, r"Encoder $q_\phi(z|...)$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, "Latent Task $z$", bbox=dict(boxstyle="circle", fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_healthcare_rl_pipeline(ax):
|
| """Applied: Healthcare / Medical Therapy."""
|
| ax.axis('off')
|
| ax.set_title("Medical RL Therapy Pipeline", fontsize=12, fontweight='bold')
|
| blocks = ["Patient History (EHR)", "State Estimator", "Policy (Action = Dose)", "Clinical Outcome"]
|
| for i, b in enumerate(blocks):
|
| ax.text(0.5, 0.9 - i*0.25, b, bbox=dict(fc="pink", alpha=0.3), ha='center')
|
| if i < 3: ax.annotate("", xy=(0.5, 0.7 - i*0.25), xytext=(0.5, 0.8 - i*0.25), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_supply_chain_rl(ax):
|
| """Applied: Supply Chain / Inventory RL."""
|
| ax.axis('off')
|
| ax.set_title("Supply Chain RL Pipeline", fontsize=12, fontweight='bold')
|
| G = nx.DiGraph()
|
| nodes = ["Factory", "Warehouse", "Retailer", "Customer"]
|
| for i, n in enumerate(nodes):
|
| ax.text(0.1 + i*0.27, 0.5, n, bbox=dict(boxstyle="round", fc="ivory"), ha='center')
|
| for i in range(3):
|
| ax.annotate("", xy=(0.28 + i*0.27, 0.5), xytext=(0.2 + i*0.27, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.text(0.5, 0.2, "State = Stock Levels, Action = Orders", ha='center', fontsize=8)
|
|
|
| def plot_sysid_safe_loop(ax):
|
| """Robotics: Sim-to-Real SysID Loop."""
|
| ax.axis('off')
|
| ax.set_title("Sim-to-Real SysID Loop", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Physical System", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.5, "System ID Estimator", bbox=dict(fc="orange", alpha=0.5), ha='center')
|
| ax.text(0.5, 0.2, "Simulation Model", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate("Observables", xy=(0.4, 0.6), xytext=(0.4, 0.75), arrowprops=dict(arrowstyle="<-"))
|
| ax.annotate("Update Parameters", xy=(0.6, 0.3), xytext=(0.6, 0.45), arrowprops=dict(arrowstyle="<-"))
|
|
|
| def plot_transformer_world_model(ax):
|
| """Architecture: Transformer World Model."""
|
| ax.axis('off')
|
| ax.set_title("Transformer World Model", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Sequence of $(s, a, r)$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, "Self-Attention Layers", bbox=dict(fc="purple", alpha=0.3), ha='center')
|
| ax.text(0.5, 0.2, "Predicted $s_{t+1}, r_{t+1}$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_network_rl(ax):
|
| """Applied: RL for Networking."""
|
| ax.axis('off')
|
| ax.set_title("Network Traffic RL", fontsize=12, fontweight='bold')
|
| G = nx.Graph()
|
| G.add_edges_from([(0,1), (1,2), (2,3), (3,0)])
|
| pos = nx.spring_layout(G)
|
| nx.draw(G, pos, ax=ax, node_color='lightblue', with_labels=False)
|
| ax.annotate("RL Router", xy=(pos[1][0], pos[1][1]), xytext=(pos[1][0], pos[1][1]+0.2), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_rlhf_ppo_ref(ax):
|
| """Training: RLHF PPO with Reference Policy."""
|
| ax.axis('off')
|
| ax.set_title("RLHF: PPO with Reference Policy", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.8, r"Active Policy $\pi_\theta$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.7, 0.8, r"Ref Policy $\pi_{ref}$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.5, 0.5, "KL Penalty", bbox=dict(boxstyle="sawtooth", fc="red", alpha=0.3), ha='center')
|
| ax.text(0.5, 0.2, "Reward Model $r(s,a)$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.6), xytext=(0.3, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.6, 0.6), xytext=(0.7, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Total Reward", xy=(0.5, 0.4), xytext=(0.5, 0.3), arrowprops=dict(arrowstyle="<-"))
|
|
|
| def plot_psro_meta_game(ax):
|
| """Multi-Agent: PSRO Meta-Game Tree."""
|
| ax.axis('off')
|
| ax.set_title("PSRO Meta-Game Update", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Meta-Game Matrix", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.2, 0.5, "Best Response", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, "Nash Equilibrium", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.5, 0.2, "Add Oracle Policy", bbox=dict(fc="gold"), ha='center', fontweight='bold')
|
| ax.annotate("", xy=(0.3, 0.6), xytext=(0.45, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.7, 0.6), xytext=(0.55, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.3, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_dial_comm_channel(ax):
|
| """Multi-Agent: DIAL Comm Channel."""
|
| ax.axis('off')
|
| ax.set_title("DIAL: Differentiable Comm", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Agent 1", bbox=dict(boxstyle="circle", fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, "Agent 2", bbox=dict(boxstyle="circle", fc="lightblue"), ha='center')
|
| ax.annotate("Message $m$ (Differentiable)", xy=(0.7, 0.52), xytext=(0.3, 0.52), arrowprops=dict(arrowstyle="->", lw=2, color='orange'))
|
| ax.annotate("Gradient $\\nabla m$", xy=(0.3, 0.48), xytext=(0.7, 0.48), arrowprops=dict(arrowstyle="->", lw=1, color='blue', ls='--'))
|
|
|
| def plot_fqi_batch_loop(ax):
|
| """Batch RL: Fitted Q-Iteration (FQI)."""
|
| ax.axis('off')
|
| ax.set_title("Fitted Q-Iteration Loop", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"Dataset $\mathcal{D}$", bbox=dict(boxstyle="round", fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, "Supervised Regressor", bbox=dict(fc="orange", alpha=0.3), ha='center')
|
| ax.text(0.5, 0.2, "Updated $Q_{k+1}$", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Bootstrap", xy=(0.8, 0.3), xytext=(0.8, 0.7), arrowprops=dict(arrowstyle="<-", connectionstyle="arc3,rad=-0.5"))
|
|
|
| def plot_cmdp_feasible_set(ax):
|
| """Safety RL: CMDP Feasible Set."""
|
| ax.set_title("CMDP Feasible Region", fontsize=12, fontweight='bold')
|
| circle = plt.Circle((0, 0), 1, alpha=0.2, color='green', label="Constrained Feasible Set")
|
| ax.add_patch(circle)
|
| ax.axhline(0.7, color='red', ls='--', label=r"Constraint $J \leq C$")
|
| ax.text(0, -0.3, r"Optimized Policy $\pi^*$", color='blue', fontweight='bold', ha='center')
|
| ax.set_xlim(-1.5, 1.5); ax.set_ylim(-1.5, 1.5)
|
| ax.legend(fontsize=8)
|
|
|
| def plot_mpc_vs_rl_horizon(ax):
|
| """Control: MPC vs RL Comparison."""
|
| ax.axis('off')
|
| ax.set_title("MPC vs RL Planning", fontsize=12, fontweight='bold')
|
| ax.text(0.25, 0.8, "MPC", fontweight='bold')
|
| ax.text(0.75, 0.8, "RL", fontweight='bold')
|
| ax.text(0.25, 0.5, "Receding Horizon\nPlanning at every step", ha='center', fontsize=8)
|
| ax.text(0.75, 0.5, "Direct Mapping from\nState to Action (Policy)", ha='center', fontsize=8)
|
| ax.text(0.5, 0.2, "Convergent when Model is Exact", color='grey', ha='center', fontsize=7)
|
|
|
| def plot_l2o_meta_pipeline(ax):
|
| """AutoML: Learning to Optimize (L2O)."""
|
| ax.axis('off')
|
| ax.set_title("Learning to Optimize (L2O)", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.7, "Optimizer (RL Policy)", bbox=dict(fc="cornflowerblue"), ha='center')
|
| ax.text(0.5, 0.3, "Optimizee (Deep Net)", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate(r"Step $\Delta w$", xy=(0.5, 0.4), xytext=(0.5, 0.6), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate(r"Gradient $\nabla L$", xy=(0.2, 0.6), xytext=(0.2, 0.4), arrowprops=dict(arrowstyle="->", color='red'))
|
|
|
| def plot_chip_placement_rl(ax):
|
| """Applied: RL for Chip Placement."""
|
| ax.set_title("RL for Chip Placement", fontsize=12, fontweight='bold')
|
| ax.grid(True, ls='--', alpha=0.3)
|
| for _ in range(8):
|
| pos = np.random.rand(2)
|
| rect = plt.Rectangle(pos, 0.1, 0.1, facecolor='lightblue', edgecolor='blue', alpha=0.7)
|
| ax.add_patch(rect)
|
| ax.set_xlim(0, 1); ax.set_ylim(0, 1)
|
| ax.text(0.5, -0.15, "Optimizing Macro Placement on Silicon", ha='center', fontsize=8)
|
|
|
| def plot_compiler_mlgo(ax):
|
| """Applied: RL for Compiler Optimization (MLGO)."""
|
| ax.axis('off')
|
| ax.set_title("MLGO: Compiler RL", fontsize=12, fontweight='bold')
|
| G = nx.DiGraph()
|
| G.add_edges_from([(0,1), (0,2), (1,3), (2,3)])
|
| pos = {0: (0.5, 0.9), 1: (0.3, 0.6), 2: (0.7, 0.6), 3: (0.5, 0.3)}
|
| nx.draw(G, pos, ax=ax, node_color='lightgreen', with_labels=False)
|
| ax.text(0.5, 0.1, "Control Flow Graph (CFG) + Inline Policy", ha='center', fontsize=8)
|
|
|
| def plot_theorem_proving_rl(ax):
|
| """Applied: RL for Theorem Proving."""
|
| ax.axis('off')
|
| ax.set_title("RL for Theorem Proving", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.9, "Target Theorem", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.3, 0.5, "Proof Step $a$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.7, 0.5, "Heuristic $V(s)$", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.2, "Verified Proof Tree", ha='center', fontsize=8)
|
| ax.annotate("", xy=(0.35, 0.6), xytext=(0.45, 0.8), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.65, 0.6), xytext=(0.55, 0.8), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_diffusion_ql_loop(ax):
|
| """Modern: Diffusion-QL Offline RL."""
|
| ax.axis('off')
|
| ax.set_title("Diffusion-QL Training", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"Noise $\epsilon$", ha='center')
|
| ax.text(0.5, 0.5, r"Denoising MLP\n$\pi_\theta(a|s, k)$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.8, 0.5, "Action $a$", ha='center')
|
| ax.annotate("", xy=(0.35, 0.5), xytext=(0.25, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.65, 0.5), xytext=(1.0, 0.5), arrowprops=dict(arrowstyle="<-"))
|
| ax.text(0.5, 0.2, "Policy as a Reverse Diffusion Process", fontsize=8, ha='center')
|
|
|
| def plot_fairness_rl_pareto(ax):
|
| """Principles: Fairness-aware RL Pareto."""
|
| ax.set_title("Fairness-Reward Pareto Frontier", fontsize=12, fontweight='bold')
|
| x = np.linspace(0.1, 1, 100)
|
| y = 1 - x**2
|
| ax.plot(x, y, color='purple', lw=3, label="Pareto Frontier")
|
| ax.fill_between(x, 0, y, color='purple', alpha=0.1)
|
| ax.set_xlabel("Reward $R$"); ax.set_ylabel("Fairness Metric $F$")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_dp_rl_noise(ax):
|
| """Principles: Differentially Private RL."""
|
| ax.axis('off')
|
| ax.set_title("Differentially Private RL", fontsize=12, fontweight='bold')
|
| ax.text(0.3, 0.5, r"Algorithm $\mathcal{A}$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, r"$\mathcal{N}(0, \sigma^2 \mathbb{I})$", bbox=dict(fc="red", alpha=0.3), ha='center')
|
| ax.text(0.7, 0.5, r"Privacy Budget $\epsilon, \delta$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.6, 0.5), xytext=(0.45, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_smart_agriculture_rl(ax):
|
| """Applied: Smart Agriculture RL."""
|
| ax.axis('off')
|
| ax.set_title("Smart Agriculture RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Soil/Weather Sensors", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.5, 0.5, "Irrigation Policy", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.2, "Yield Optimization", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_climate_rl_grid(ax):
|
| """Applied: Climate Science RL."""
|
| ax.set_title("Climate Mitigation RL (Grid)", fontsize=12, fontweight='bold')
|
| data = np.random.randn(10, 10)
|
| im = ax.imshow(data, cmap='coolwarm')
|
| ax.set_xlabel("Longitude"); ax.set_ylabel("Latitude")
|
| ax.text(5, 5, "Carbon Sequestration\nControl Map", ha='center', color='white', fontweight='bold', fontsize=8)
|
|
|
| def plot_ai_education_tracing(ax):
|
| """Applied: Intelligent Tutoring Systems RL."""
|
| ax.axis('off')
|
| ax.set_title("AI Education (Knowledge Tracing)", fontsize=12, fontweight='bold')
|
| nodes = ["Concept 1", "Concept 2", "Student State $S_t$", "Next Problem $a_t$"]
|
| for i, n in enumerate(nodes):
|
| ax.text(0.2 + (i%2)*0.6, 0.7 - (i//2)*0.4, n, bbox=dict(fc="pink", alpha=0.3), ha='center')
|
| ax.annotate("", xy=(0.6, 0.5), xytext=(0.4, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_decision_sde_flow(ax):
|
| """Modern: Decision SDEs."""
|
| ax.set_title(r"Decision SDE Flow $dX_t = f(X_t, u_t)dt + gdW_t$", fontsize=10, fontweight='bold')
|
| t = np.linspace(0, 1, 100)
|
| for _ in range(5):
|
| path = np.cumsum(np.random.normal(0, 0.1, size=100))
|
| ax.plot(t, path + 0.5*t, alpha=0.5)
|
| ax.set_xlabel("Continuous Time $t$")
|
|
|
| def plot_diff_physics_brax(ax):
|
| """Control: Differentiable Physics (Brax)."""
|
| ax.axis('off')
|
| ax.set_title(r"Differentiable physics $\nabla_{u} \mathcal{L}$", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Physics Engine (Jacobian)", bbox=dict(fc="orange", alpha=0.1), ha='center')
|
| ax.text(0.5, 0.5, "Simulator Layer", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.5, 0.2, "Policy Update", bbox=dict(fc="blue", alpha=0.1), ha='center')
|
| ax.annotate("", xy=(0.5, 0.4), xytext=(0.5, 0.6), arrowprops=dict(arrowstyle="<-", color='red', label="Grads"))
|
|
|
| def plot_beamforming_rl(ax):
|
| """Applied: RL for Beamforming."""
|
| ax.axis('off')
|
| ax.set_title("Wireless Beamforming RL", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.2, 0.5), 0.05, color='black'))
|
| theta = np.linspace(-np.pi/4, np.pi/4, 100)
|
| r = np.cos(4*theta)
|
| ax.plot(0.2 + r*np.cos(theta), 0.5 + r*np.sin(theta), color='orange', label="Main Lobe")
|
| ax.text(0.8, 0.5, "User Device", bbox=dict(boxstyle="round", fc="lightgrey"), ha='center')
|
|
|
| def plot_quantum_error_correction_rl(ax):
|
| """Applied: Quantum Error Correction RL."""
|
| ax.axis('off')
|
| ax.set_title("Quantum Error Correction RL", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Syndrome $S$", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, "Decoder Agent", bbox=dict(boxstyle="round4", fc="purple", alpha=0.2), ha='center')
|
| ax.text(0.9, 0.5, "Recovery $P$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_mean_field_rl(ax):
|
| """Multi-Agent: Mean Field RL."""
|
| ax.axis('off')
|
| ax.set_title("Mean Field RL Interaction", fontsize=12, fontweight='bold')
|
| x = np.random.randn(50)
|
| ax.text(0.2, 0.5, "Single Agent $i$", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, r"Mean State $\overline{s}$", bbox=dict(fc="white"), ha='center', fontweight='bold')
|
| ax.annotate("", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="<->"))
|
| ax.text(0.5, 0.2, r"Population Limit $N \rightarrow \infty$", ha='center', fontsize=8)
|
|
|
| def plot_goal_gan_hrl(ax):
|
| """HRL: Goal-GAN Pipeline."""
|
| ax.axis('off')
|
| ax.set_title("Goal-GAN Curriculum", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.7, "Goal Generator\n(GAN Ref)", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.8, 0.7, "RL Policy\n(Worker)", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.5, 0.3, "Goal Label (Success/Fail)", bbox=dict(fc="ivory"), ha='center')
|
| ax.annotate("Set Goal $g$", xy=(0.7, 0.7), xytext=(0.3, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("Train GAN", xy=(0.3, 0.4), xytext=(0.5, 0.35), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_jepa_arch(ax):
|
| """Modern: JEPA (Joint Embedding Predictive Architecture)."""
|
| ax.axis('off')
|
| ax.set_title("JEPA: Predictive Architecture", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.2, "Context $x$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.8, 0.2, "Target $y$", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.2, 0.6, "Encoder $E_x$", bbox=dict(fc="cornflowerblue"), ha='center')
|
| ax.text(0.8, 0.6, "Encoder $E_y$", bbox=dict(fc="cornflowerblue"), ha='center')
|
| ax.text(0.5, 0.8, "Predictor $P$", bbox=dict(fc="orange", alpha=0.3), ha='center')
|
| for i in [0.2, 0.8]:
|
| ax.annotate("", xy=(i, 0.5), xytext=(i, 0.3), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.4, 0.75), xytext=(0.25, 0.65), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.6, 0.75), xytext=(0.75, 0.65), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_cql_penalty_surface(ax):
|
| """Offline RL: CQL Value Penalty."""
|
| X, Y = np.meshgrid(np.linspace(-3, 3, 20), np.linspace(-3, 3, 20))
|
| Z = (X**2 + Y**2) - 2 * np.exp(- (X**2 + Y**2))
|
| ax.contourf(X, Y, Z, cmap='viridis')
|
| ax.set_title("CQL Value Penalty Landscape", fontsize=12, fontweight='bold')
|
|
|
| def plot_cyber_attack_defense(ax):
|
| """Applied: Cybersecurity RL Game."""
|
| ax.axis('off')
|
| ax.set_title("Cybersecurity Attack-Defense RL", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.7, "Attacker Agent", bbox=dict(fc="red", alpha=0.2), ha='center', fontweight='bold')
|
| ax.text(0.8, 0.7, "Defender Agent", bbox=dict(fc="blue", alpha=0.2), ha='center', fontweight='bold')
|
| ax.text(0.5, 0.3, "Network Infrastructure", bbox=dict(fc="grey", alpha=0.3), ha='center')
|
| ax.annotate("Intrusion", xy=(0.4, 0.4), xytext=(0.2, 0.6), arrowprops=dict(arrowstyle="->", color='red'))
|
| ax.annotate("Mitigation", xy=(0.6, 0.4), xytext=(0.8, 0.6), arrowprops=dict(arrowstyle="->", color='blue'))
|
|
|
| def plot_causal_irl(ax):
|
| """Causal: Causal Inverse RL Graph."""
|
| ax.axis('off')
|
| ax.set_title("Causal Inverse RL Graph", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.8, "State $S$", ha='center', bbox=dict(fc="ivory"))
|
| ax.text(0.5, 0.8, "Latent Factor $U$", ha='center', bbox=dict(fc="red", alpha=0.1), fontweight='bold')
|
| ax.text(0.2, 0.4, "Action $A$", ha='center', bbox=dict(fc="lightblue"))
|
| ax.text(0.5, 0.4, "Reward $R$", ha='center', bbox=dict(fc="gold"))
|
| ax.annotate("", xy=(0.2, 0.5), xytext=(0.2, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.5, 0.7), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.3, 0.4), xytext=(0.4, 0.4), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_vqe_rl(ax):
|
| """Quantum: VQE-RL Circuit Optimization."""
|
| ax.axis('off')
|
| ax.set_title("VQE-RL Optimization", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Quantum Circuit $U(\\theta)$", bbox=dict(fc="purple", alpha=0.1), ha='center')
|
| ax.text(0.5, 0.5, "Energy Expectation $\\langle H \\rangle$", ha='center')
|
| ax.text(0.5, 0.2, "RL Optimizer", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_drug_discovery_rl(ax):
|
| """Applied: RL for Drug Discovery."""
|
| ax.axis('off')
|
| ax.set_title("De-novo Drug Discovery RL", fontsize=12, fontweight='bold')
|
| ax.text(0.1, 0.5, "Seed Molecule", ha='center')
|
| ax.text(0.5, 0.5, "RL Modification Step", bbox=dict(fc="green", alpha=0.1), ha='center')
|
| ax.text(0.9, 0.5, "Optimized Lead", ha='center')
|
| ax.annotate("", xy=(0.4, 0.5), xytext=(0.2, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.8, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_traffic_signal_coordination(ax):
|
| """Applied: Traffic Signal RL."""
|
| ax.axis('off')
|
| ax.set_title("Traffic Signal Coordination RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Intersection Grid", ha='center')
|
| ax.text(0.2, 0.5, "Signal A (RL)", bbox=dict(fc="red"), ha='center')
|
| ax.text(0.8, 0.5, "Signal B (RL)", bbox=dict(fc="green"), ha='center')
|
| ax.annotate("Max-Pressure Reward", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="<->", color='orange'))
|
|
|
| def plot_mars_rover_pathfinding(ax):
|
| """Applied: Mars Rover RL."""
|
| ax.set_title("Mars Rover Pathfinding RL", fontsize=12, fontweight='bold')
|
| x = np.linspace(0, 5, 20)
|
| y = np.sin(x) + np.cos(x*0.5)
|
| ax.plot(x, y, color='brown', lw=2, label="Terrain")
|
| ax.scatter([1, 4], [y[4], y[16]], color='red', label="Waypoints")
|
| ax.legend(fontsize=8)
|
|
|
| def plot_sports_analytics_rl(ax):
|
| """Applied: Sports Analytics RL."""
|
| ax.set_title("Sports Player Movement RL", fontsize=12, fontweight='bold')
|
| x = np.random.normal(0, 1, 50)
|
| y = np.random.normal(0, 1, 50)
|
| ax.hexbin(x, y, gridsize=10, cmap='Blues')
|
| ax.text(0, 0, "High Pressure Zone", ha='center', color='black', alpha=0.5)
|
|
|
| def plot_crypto_attack_rl(ax):
|
| """Applied: Cryptography Attack RL."""
|
| ax.axis('off')
|
| ax.set_title("Cryptography Attack RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Cipher State", ha='center')
|
| ax.text(0.5, 0.5, "Differential Cryptanalysis Search", bbox=dict(fc="red", alpha=0.1), ha='center')
|
| ax.text(0.5, 0.2, "Broken Key Found", ha='center', fontweight='bold')
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_humanitarian_rl(ax):
|
| """Applied: Humanitarian Aid RL."""
|
| ax.axis('off')
|
| ax.set_title("Humanitarian Resource RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Disaster Zone Clusters", ha='center')
|
| ax.text(0.2, 0.4, "Supply Hub", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.text(0.8, 0.4, "Isolated Community", bbox=dict(fc="red", alpha=0.1), ha='center')
|
| ax.annotate("Optimal Cargo Drop", xy=(0.7, 0.4), xytext=(0.3, 0.4), arrowprops=dict(arrowstyle="->", lw=2, color='blue'))
|
|
|
| def plot_video_compression_rl(ax):
|
| """Applied: Video Compression RL."""
|
| ax.set_title("Video Compression RL (Rate-Distortion)", fontsize=12, fontweight='bold')
|
| bitrate = np.logspace(0, 10, 100)
|
| distortion = 1 / bitrate
|
| ax.loglog(bitrate, distortion, label=r"Policy $\pi$ RD curve")
|
| ax.set_xlabel("Bit Rate"); ax.set_ylabel("Distortion")
|
| ax.legend()
|
|
|
| def plot_kubernetes_scaling_rl(ax):
|
| """Applied: Kubernetes Scaling RL."""
|
| ax.axis('off')
|
| ax.set_title("Kubernetes Auto-scaling RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Cloud Service Load", ha='center')
|
| ax.text(0.5, 0.5, "RL Autoscaler", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.2, "Replicas count $n$", ha='center')
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_fluid_dynamics_rl(ax):
|
| """Applied: Fluid Dynamics Control RL."""
|
| ax.set_title("Fluid Dynamics Flow Control RL", fontsize=12, fontweight='bold')
|
| Y, X = np.mgrid[-1:1:100j, -1:1:100j]
|
| U = -1 - X**2 + Y
|
| V = 1 + X - Y**2
|
| ax.streamplot(X, Y, U, V, color='cornflowerblue')
|
| ax.set_title("Flow Control optimization")
|
|
|
| def plot_structural_optimization_rl(ax):
|
| """Applied: Structural RL."""
|
| ax.axis('off')
|
| ax.set_title("Structural Optimization RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Bridge Topology State", ha='center')
|
| ax.text(0.5, 0.5, "Agent Stress Estimation", ha='center')
|
| ax.text(0.5, 0.2, "Material Placement Action", ha='center', bbox=dict(fc="lightgrey"))
|
|
|
| def plot_human_decision_rl(ax):
|
| """Applied: Human Modeling RL."""
|
| ax.set_title("Human Decision Modeling (Prospect Theory)", fontsize=12, fontweight='bold')
|
| x = np.linspace(-10, 10, 100)
|
| y = np.where(x > 0, x**0.88, -2.25*(-x)**0.88)
|
| ax.plot(x, y, label="Human Value Function")
|
| ax.axvline(0, color='black', alpha=0.2)
|
| ax.legend()
|
|
|
| def plot_semantic_parsing_rl(ax):
|
| """Applied: Semantic Parsing RL."""
|
| ax.axis('off')
|
| ax.set_title("Semantic Parsing RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Natural Language Input", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.5, "Logic derivation step $a$", ha='center')
|
| ax.text(0.5, 0.2, "SQL/Lambda Expr Tree", ha='center', fontweight='bold')
|
|
|
| def plot_music_melody_rl(ax):
|
| """Applied: Music Composition RL."""
|
| ax.axis('off')
|
| ax.set_title("Melody generation RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.5, "Sequence of Notes\n(C, E, G, B...)", bbox=dict(fc="ivory"), ha='center')
|
| ax.text(0.5, 0.2, "Aesthetic Reward Model", bbox=dict(fc="pink"), ha='center')
|
|
|
| def plot_plasma_control_rl(ax):
|
| """Applied: Plasma Fusion RL."""
|
| ax.axis('off')
|
| ax.set_title("Plasma Fusion Control RL", fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.3, color='orange', alpha=0.5, label="Plasma"))
|
| ax.text(0.5, 0.5, "Tokamak Center", ha='center')
|
| ax.annotate("Magnetic Coil Action", xy=(0.8, 0.8), xytext=(0.6, 0.6), arrowprops=dict(arrowstyle="->", color='red'))
|
|
|
| def plot_carbon_capture_rl(ax):
|
| """Applied: Carbon Capture RL."""
|
| ax.axis('off')
|
| ax.set_title("Carbon Capture RL cycle", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Adsorption", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.text(0.8, 0.5, "Desorption", bbox=dict(fc="orange"), ha='center')
|
| ax.annotate("", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.3, 0.45), xytext=(0.7, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_swarm_robotics_rl(ax):
|
| """Applied: Swarm RL."""
|
| ax.axis('off')
|
| ax.set_title("Swarm Robotics RL", fontsize=12, fontweight='bold')
|
| for _ in range(10):
|
| pos = np.random.rand(2)
|
| ax.add_patch(plt.Circle(pos, 0.02, color='black'))
|
| ax.text(0.5, 1, "Emergent Coordination Plan", ha='center')
|
|
|
| def plot_legal_compliance_rl(ax):
|
| """Applied: Legal RL."""
|
| ax.axis('off')
|
| ax.set_title("Legal Compliance RL Game", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, r"Regulation $\mathcal{L}$", ha='center')
|
| ax.text(0.8, 0.5, r"Compliance Policy $\pi$", ha='center', bbox=dict(fc="gold"))
|
| ax.annotate("Audit", xy=(0.2, 0.4), xytext=(0.8, 0.4), arrowprops=dict(arrowstyle="->", ls='--'))
|
|
|
| def plot_pinn_rl_loss(ax):
|
| """Physics: Physics-Informed RL (PINN)."""
|
| ax.axis('off')
|
| ax.set_title("Physics-Informed RL (PINN)", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, r"RL Loss $\mathcal{L}_{RL}$", ha='center')
|
| ax.text(0.5, 0.5, r"PDE Constraint $\mathcal{L}_{Phys}$", ha='center', color='red')
|
| ax.text(0.5, 0.2, r"Optimized Policy $\pi_\theta$", fontweight='bold', ha='center')
|
|
|
| def plot_neuro_symbolic_rl(ax):
|
| """Modern: Neuro-Symbolic RL."""
|
| ax.axis('off')
|
| ax.set_title("Neuro-Symbolic RL", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.5, "Neural State", bbox=dict(fc="lightblue"), ha='center')
|
| ax.text(0.8, 0.5, "Symbolic Logic", bbox=dict(fc="lightgreen"), ha='center')
|
| ax.annotate("Abstraction", xy=(0.7, 0.5), xytext=(0.3, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_defi_liquidity_rl(ax):
|
| """Applied: DeFi RL."""
|
| ax.axis('off')
|
| ax.set_title("DeFi Liquidity Pool RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.7, "Liquidity Pool $(x, y)$", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.3, "LP Strategy Policy", ha='center')
|
| ax.annotate("Arbitrage Action", xy=(1.0, 0.5), xytext=(0.6, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_dopamine_rpe_curves(ax):
|
| """Neuroscience: Dopamine RPE."""
|
| ax.set_title("Dopamine Reward Prediction Error", fontsize=12, fontweight='bold')
|
| t = np.linspace(0, 10, 100)
|
| rpe = np.exp(-t)
|
| ax.plot(t, rpe, label=r"Expected RPE $\delta$")
|
| ax.set_ylabel("Dopamine neurons firing rate")
|
| ax.legend()
|
|
|
| def plot_proprioceptive_rl_loop(ax):
|
| """Robotics: Proprioceptive RL."""
|
| ax.axis('off')
|
| ax.set_title("Proprioceptive Sensory-Motor RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Joint Encoders", ha='center')
|
| ax.text(0.5, 0.4, "Low-level Controller", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.5), xytext=(0.5, 0.7), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_ar_placement_rl(ax):
|
| """Applied: AR RL."""
|
| ax.axis('off')
|
| ax.set_title("Augmented Reality Object Placement RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.5, "[AR View of Room]", ha='center')
|
| ax.text(0.8, 0.8, "Optimal overlay position", ha='center', color='blue')
|
|
|
| def plot_sequential_bundle_rl(ax):
|
| """Recommendation: Sequential Bundle RL."""
|
| ax.axis('off')
|
| ax.set_title("Sequential Bundle RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "User context sequence", ha='center')
|
| nodes = ["Item 1", "Item 2", "Item 3"]
|
| for i, n in enumerate(nodes):
|
| ax.text(0.2 + i*0.3, 0.5, n, bbox=dict(fc="ivory"), ha='center')
|
|
|
| def plot_ogd_vs_rl_gradient(ax):
|
| """Theoretical: OGD vs RL."""
|
| ax.set_title("Online Gradient Descent vs RL", fontsize=12, fontweight='bold')
|
| x = np.linspace(-3, 3, 100)
|
| ax.plot(x, x**2, label="OGD loss curve")
|
| ax.plot(x, -np.log(1/(1+np.exp(-x))), label="RL surrogate loss", ls='--')
|
| ax.legend()
|
|
|
| def plot_active_learning_selection(ax):
|
| """Modern: Active Learning RL."""
|
| ax.axis('off')
|
| ax.set_title("Active Learning: Query RL Selection", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Pool of unlabeled samples", ha='center')
|
| ax.text(0.5, 0.5, r"Acquisition Policy $\pi$", bbox=dict(fc="gold"), ha='center')
|
| ax.annotate("Send to Oracle", xy=(1.0, 0.5), xytext=(0.7, 0.5), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_federated_rl_tree(ax):
|
| """Modern: Federated RL."""
|
| ax.axis('off')
|
| ax.set_title("Federated RL global Aggregator", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Global Model / Server", bbox=dict(fc="purple", alpha=0.1), ha='center')
|
| for i in [0.2, 0.5, 0.8]:
|
| ax.text(i, 0.4, f"Local Agent {int(i*3)}", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.75), xytext=(i, 0.45), arrowprops=dict(arrowstyle="<->"))
|
|
|
| def plot_ultimate_mastery_diagram(ax):
|
| """Conceptual: Ultimate Universal RL Mastery Diagram."""
|
| ax.axis('off')
|
| ax.set_title("UTIMATE UNIVERSAL RL MASTERY MILESTONE (230)", fontsize=16, fontweight='bold', color='darkred')
|
| ax.text(0.5, 0.5, "The Definitive\nMaster Anthology\nof Reinforcement Learning", ha='center', fontsize=12, fontweight='bold')
|
| ax.add_patch(plt.Circle((0.5, 0.5), 0.4, color='gold', alpha=0.05, lw=5, edgecolor='black'))
|
| ax.text(0.5, 0.1, "230 UNIQUE GRAPHICAL REPRESENTATIONS ACHIEVED", ha='center', fontweight='bold')
|
|
|
|
|
| def plot_smart_grid_rl(ax):
|
| """Applied: Smart Grid Supply/Demand."""
|
| ax.axis('off')
|
| ax.set_title("Smart Grid RL Management", fontsize=12, fontweight='bold')
|
| ax.text(0.2, 0.8, "Renewables", ha='center')
|
| ax.text(0.8, 0.8, "Consumers", ha='center')
|
| ax.text(0.5, 0.5, "RL Dispatcher", bbox=dict(fc="gold"), ha='center')
|
| ax.text(0.5, 0.2, "Energy Storage", bbox=dict(fc="lightgrey"), ha='center')
|
| ax.annotate("", xy=(0.4, 0.55), xytext=(0.25, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.75, 0.75), xytext=(0.6, 0.6), arrowprops=dict(arrowstyle="<-"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_quantum_tomography_rl(ax):
|
| """Applied: Quantum State Tomography."""
|
| ax.axis('off')
|
| ax.set_title("Quantum state Tomography RL", fontsize=12, fontweight='bold')
|
| ax.text(0.5, 0.8, "Quantum State $\\rho$", bbox=dict(boxstyle="circle", fc="purple", alpha=0.2), ha='center')
|
| ax.text(0.5, 0.5, "Measurement $M$", ha='center')
|
| ax.text(0.5, 0.2, "RL Estimator", bbox=dict(fc="lightblue"), ha='center')
|
| ax.annotate("", xy=(0.5, 0.6), xytext=(0.5, 0.75), arrowprops=dict(arrowstyle="->"))
|
| ax.annotate("", xy=(0.5, 0.3), xytext=(0.5, 0.45), arrowprops=dict(arrowstyle="->"))
|
|
|
| def plot_absolute_encyclopedia_map(ax):
|
| """Conceptual: Absolute Universal Encyclopedia Map."""
|
| ax.axis('off')
|
| ax.set_title("Absolute Universal RL Pillar Map", fontsize=14, fontweight='bold', color='darkblue')
|
| categories = ["Foundational", "Model-Free", "Model-Based", "Advanced Paradigms", "Analysis/Safety", "Applied Pipelines"]
|
| for i, c in enumerate(categories):
|
| angle = 2 * np.pi * i / 6
|
| ax.text(0.5 + 0.35*np.cos(angle), 0.5 + 0.35*np.sin(angle), c, bbox=dict(fc="ivory", lw=2), ha='center', fontsize=9)
|
| ax.text(0.5, 0.5, "Reinforcement\nLearning\nGraphical\nLibrary", ha='center', fontweight='bold', fontsize=12)
|
| for i in range(6):
|
| angle = 2 * np.pi * i / 6
|
| ax.annotate("", xy=(0.5 + 0.25*np.cos(angle), 0.5 + 0.25*np.sin(angle)), xytext=(0.5, 0.5), arrowprops=dict(arrowstyle="->", alpha=0.3))
|
|
|
| def plot_actor_critic_arch(ax):
|
| """Actor-Critic: Three-network diagram (TD3 - actor + two critics)."""
|
| ax.axis('off')
|
| ax.set_title("TD3 Architecture Diagram", fontsize=12, fontweight='bold')
|
|
|
|
|
| ax.text(0.1, 0.5, r"State" + "\n" + r"$s$", ha="center", va="center", bbox=dict(boxstyle="circle,pad=0.5", fc="lightblue"))
|
|
|
|
|
| net_props = dict(boxstyle="square,pad=0.8", fc="lightgreen", ec="black")
|
| ax.text(0.5, 0.8, r"Actor $\pi_\phi$", ha="center", va="center", bbox=net_props)
|
| ax.text(0.5, 0.5, r"Critic 1 $Q_{\theta_1}$", ha="center", va="center", bbox=net_props)
|
| ax.text(0.5, 0.2, r"Critic 2 $Q_{\theta_2}$", ha="center", va="center", bbox=net_props)
|
|
|
|
|
| ax.text(0.8, 0.8, "Action $a$", ha="center", va="center", bbox=dict(boxstyle="circle,pad=0.3", fc="coral"))
|
| ax.text(0.8, 0.35, "Min Q-value", ha="center", va="center", bbox=dict(boxstyle="round,pad=0.3", fc="gold"))
|
|
|
|
|
| kwargs = dict(arrowstyle="->", lw=1.5)
|
| ax.annotate("", xy=(0.38, 0.8), xytext=(0.15, 0.55), arrowprops=kwargs)
|
| ax.annotate("", xy=(0.38, 0.5), xytext=(0.15, 0.5), arrowprops=kwargs)
|
| ax.annotate("", xy=(0.38, 0.2), xytext=(0.15, 0.45), arrowprops=kwargs)
|
| ax.annotate("", xy=(0.73, 0.8), xytext=(0.62, 0.8), arrowprops=kwargs)
|
| ax.annotate("", xy=(0.68, 0.35), xytext=(0.62, 0.5), arrowprops=kwargs)
|
| ax.annotate("", xy=(0.68, 0.35), xytext=(0.62, 0.2), arrowprops=kwargs)
|
|
|
| def plot_epsilon_decay(ax):
|
| """Exploration: ε-Greedy Strategy Decay Curve."""
|
| episodes = np.arange(0, 1000)
|
| epsilon = np.maximum(0.01, np.exp(-0.005 * episodes))
|
|
|
| ax.plot(episodes, epsilon, color='purple', lw=2)
|
| ax.set_title(r"$\epsilon$-Greedy Decay Curve", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Episodes")
|
| ax.set_ylabel(r"Probability $\epsilon$")
|
| ax.grid(True, linestyle='--', alpha=0.6)
|
| ax.fill_between(episodes, epsilon, color='purple', alpha=0.1)
|
|
|
| def plot_learning_curve(ax):
|
| """Advanced / Misc: Learning Curve with Confidence Bands."""
|
| steps = np.linspace(0, 1e6, 100)
|
|
|
| mean_return = 100 * (1 - np.exp(-5e-6 * steps)) + np.random.normal(0, 2, len(steps))
|
| std_dev = 15 * np.exp(-2e-6 * steps)
|
|
|
| ax.plot(steps, mean_return, color='blue', lw=2, label="PPO (Mean)")
|
| ax.fill_between(steps, mean_return - std_dev, mean_return + std_dev, color='blue', alpha=0.2, label="±1 Std Dev")
|
|
|
| ax.set_title("Learning Curve (Return vs Steps)", fontsize=12, fontweight='bold')
|
| ax.set_xlabel("Environment Steps")
|
| ax.set_ylabel("Average Episodic Return")
|
| ax.legend(loc="lower right")
|
| ax.grid(True, linestyle='--', alpha=0.6)
|
|
|
| def main():
|
|
|
| fig1, gs1 = setup_figure("RL: MDP & Environment", 2, 4)
|
|
|
| plot_agent_env_loop(fig1.add_subplot(gs1[0, 0]))
|
| plot_mdp_graph(fig1.add_subplot(gs1[0, 1]))
|
| plot_trajectory(fig1.add_subplot(gs1[0, 2]))
|
| plot_continuous_space(fig1.add_subplot(gs1[0, 3]))
|
| plot_reward_landscape(fig1, gs1)
|
| plot_discount_decay(fig1.add_subplot(gs1[1, 1]))
|
|
|
|
|
|
|
|
|
|
|
| fig2, gs2 = setup_figure("RL: Value, Policy & Dynamic Programming", 2, 4)
|
| plot_value_heatmap(fig2.add_subplot(gs2[0, 0]))
|
| plot_action_value_q(fig2.add_subplot(gs2[0, 1]))
|
| plot_policy_arrows(fig2.add_subplot(gs2[0, 2]))
|
| plot_advantage_function(fig2.add_subplot(gs2[0, 3]))
|
| plot_backup_diagram(fig2.add_subplot(gs2[1, 0]))
|
| plot_policy_improvement(fig2.add_subplot(gs2[1, 1]))
|
| plot_value_iteration_backup(fig2.add_subplot(gs2[1, 2]))
|
| plot_policy_iteration_cycle(fig2.add_subplot(gs2[1, 3]))
|
|
|
|
|
|
|
|
|
| fig3, gs3 = setup_figure("RL: Monte Carlo & Temporal Difference", 2, 4)
|
| plot_mc_backup(fig3.add_subplot(gs3[0, 0]))
|
| plot_mcts(fig3.add_subplot(gs3[0, 1]))
|
| plot_importance_sampling(fig3.add_subplot(gs3[0, 2]))
|
| plot_td_backup(fig3.add_subplot(gs3[0, 3]))
|
| plot_nstep_td(fig3.add_subplot(gs3[1, 0]))
|
| plot_eligibility_traces(fig3.add_subplot(gs3[1, 1]))
|
| plot_sarsa_backup(fig3.add_subplot(gs3[1, 2]))
|
| plot_q_learning_backup(fig3.add_subplot(gs3[1, 3]))
|
|
|
|
|
|
|
|
|
| fig4, gs4 = setup_figure("RL: TD Extensions & Function Approximation", 2, 4)
|
| plot_double_q(fig4.add_subplot(gs4[0, 0]))
|
| plot_dueling_dqn(fig4.add_subplot(gs4[0, 1]))
|
| plot_prioritized_replay(fig4.add_subplot(gs4[0, 2]))
|
| plot_rainbow_dqn(fig4.add_subplot(gs4[0, 3]))
|
| plot_linear_fa(fig4.add_subplot(gs4[1, 0]))
|
| plot_nn_layers(fig4.add_subplot(gs4[1, 1]))
|
| plot_computation_graph(fig4.add_subplot(gs4[1, 2]))
|
| plot_target_network(fig4.add_subplot(gs4[1, 3]))
|
|
|
|
|
|
|
|
|
| fig5, gs5 = setup_figure("RL: Policy Gradients, Actor-Critic & Exploration", 2, 4)
|
| plot_policy_gradient_flow(fig5.add_subplot(gs5[0, 0]))
|
| plot_ppo_clip(fig5.add_subplot(gs5[0, 1]))
|
| plot_trpo_trust_region(fig5.add_subplot(gs5[0, 2]))
|
| plot_actor_critic_arch(fig5.add_subplot(gs5[0, 3]))
|
| plot_a3c_multi_worker(fig5.add_subplot(gs5[1, 0]))
|
| plot_sac_arch(fig5.add_subplot(gs5[1, 1]))
|
| plot_softmax_exploration(fig5.add_subplot(gs5[1, 2]))
|
| plot_ucb_confidence(fig5.add_subplot(gs5[1, 3]))
|
|
|
|
|
|
|
|
|
| fig6, gs6 = setup_figure("RL: Hierarchical, Model-Based & Offline", 2, 4)
|
| plot_options_framework(fig6.add_subplot(gs6[0, 0]))
|
| plot_feudal_networks(fig6.add_subplot(gs6[0, 1]))
|
| plot_world_model(fig6.add_subplot(gs6[0, 2]))
|
| plot_model_planning(fig6.add_subplot(gs6[0, 3]))
|
| plot_offline_rl(fig6.add_subplot(gs6[1, 0]))
|
| plot_cql_regularization(fig6.add_subplot(gs6[1, 1]))
|
| plot_epsilon_decay(fig6.add_subplot(gs6[1, 2]))
|
| plot_intrinsic_motivation(fig6.add_subplot(gs6[1, 3]))
|
|
|
|
|
|
|
|
|
| fig7, gs7 = setup_figure("RL: Multi-Agent, IRL & Meta-RL", 2, 4)
|
| plot_multi_agent_interaction(fig7.add_subplot(gs7[0, 0]))
|
| plot_ctde(fig7.add_subplot(gs7[0, 1]))
|
| plot_payoff_matrix(fig7.add_subplot(gs7[0, 2]))
|
| plot_irl_reward_inference(fig7.add_subplot(gs7[0, 3]))
|
| plot_gail_flow(fig7.add_subplot(gs7[1, 0]))
|
| plot_meta_rl_nested_loop(fig7.add_subplot(gs7[1, 1]))
|
| plot_task_distribution(fig7.add_subplot(gs7[1, 2]))
|
|
|
|
|
|
|
|
|
| fig8, gs8 = setup_figure("RL: Advanced & Miscellaneous", 2, 4)
|
| plot_replay_buffer(fig8.add_subplot(gs8[0, 0]))
|
| plot_state_visitation(fig8.add_subplot(gs8[0, 1]))
|
| plot_regret_curve(fig8.add_subplot(gs8[0, 2]))
|
| plot_attention_weights(fig8.add_subplot(gs8[0, 3]))
|
| plot_diffusion_policy(fig8.add_subplot(gs8[1, 0]))
|
| plot_gnn_rl(fig8.add_subplot(gs8[1, 1]))
|
| plot_latent_space(fig8.add_subplot(gs8[1, 2]))
|
| plot_convergence_log(fig8.add_subplot(gs8[1, 3]))
|
|
|
|
|
| fig9, gs9 = setup_figure("RL: Specialized & Modern (Absolute Completeness)", 3, 4)
|
|
|
| plot_rl_taxonomy_tree(fig9.add_subplot(gs9[0, 0]))
|
| plot_rl_as_inference_pgm(fig9.add_subplot(gs9[0, 1]))
|
| plot_distributional_rl_atoms(fig9.add_subplot(gs9[0, 2]))
|
| plot_her_goal_relabeling(fig9.add_subplot(gs9[0, 3]))
|
|
|
| plot_dyna_q_flow(fig9.add_subplot(gs9[1, 0]))
|
| plot_noisy_nets_parameters(fig9.add_subplot(gs9[1, 1]))
|
| plot_icm_curiosity(fig9.add_subplot(gs9[1, 2]))
|
| plot_v_trace_impala(fig9.add_subplot(gs9[1, 3]))
|
|
|
| plot_qmix_mixing_net(fig9.add_subplot(gs9[2, 0]))
|
| plot_saliency_heatmaps(fig9.add_subplot(gs9[2, 1]))
|
| plot_tsne_state_embeddings(fig9.add_subplot(gs9[2, 2]))
|
| plot_action_selection_noise(fig9.add_subplot(gs9[2, 3]))
|
|
|
|
|
| fig10, gs10 = setup_figure("RL: Evaluation, Safety & Alignment", 2, 4)
|
| plot_success_rate_curve(fig10.add_subplot(gs10[0, 0]))
|
| plot_performance_profiles_rliable(fig10.add_subplot(gs10[0, 1]))
|
| plot_hyperparameter_sensitivity(fig10.add_subplot(gs10[0, 2]))
|
| plot_action_persistence(fig10.add_subplot(gs10[0, 3]))
|
| plot_safety_shielding(fig10.add_subplot(gs10[1, 0]))
|
| plot_automated_curriculum(fig10.add_subplot(gs10[1, 1]))
|
| plot_domain_randomization(fig10.add_subplot(gs10[1, 2]))
|
| plot_rlhf_flow(fig10.add_subplot(gs10[1, 3]))
|
|
|
|
|
| fig11, gs11 = setup_figure("RL: Transformers & Specific MB Architecture", 1, 3)
|
| plot_decision_transformer_tokens(fig11.add_subplot(gs11[0, 0]))
|
| plot_muzero_search_tree(fig11.add_subplot(gs11[0, 1]))
|
| plot_policy_distillation(fig11.add_subplot(gs11[0, 2]))
|
|
|
|
|
|
|
| fig_loss = plt.figure(figsize=(10, 8))
|
| gs_loss = GridSpec(1, 1, figure=fig_loss)
|
| plot_loss_landscape(fig_loss, gs_loss)
|
|
|
| plt.show()
|
|
|
| def save_all_graphs(output_dir="graphs"):
|
| """Saves each of the 74 RL components as a separate PNG file."""
|
| if not os.path.exists(output_dir):
|
| os.makedirs(output_dir)
|
|
|
|
|
| mapping = {
|
| "Agent-Environment Interaction Loop": plot_agent_env_loop,
|
| "Markov Decision Process (MDP) Tuple": plot_mdp_graph,
|
| "State Transition Graph": plot_mdp_graph,
|
| "Trajectory / Episode Sequence": plot_trajectory,
|
| "Continuous State/Action Space Visualization": plot_continuous_space,
|
| "Reward Function / Landscape": plot_reward_landscape,
|
| "Discount Factor (gamma) Effect": plot_discount_decay,
|
| "State-Value Function V(s)": plot_value_heatmap,
|
| "Action-Value Function Q(s,a)": plot_action_value_q,
|
| "Policy pi(s) or pi(a|s)": plot_policy_arrows,
|
| "Advantage Function A(s,a)": plot_advantage_function,
|
| "Optimal Value Function V* / Q*": plot_value_heatmap,
|
| "Policy Evaluation Backup": plot_backup_diagram,
|
| "Policy Improvement": plot_policy_improvement,
|
| "Value Iteration Backup": plot_value_iteration_backup,
|
| "Policy Iteration Full Cycle": plot_policy_iteration_cycle,
|
| "Monte Carlo Backup": plot_mc_backup,
|
| "Monte Carlo Tree (MCTS)": plot_mcts,
|
| "Importance Sampling Ratio": plot_importance_sampling,
|
| "TD(0) Backup": plot_td_backup,
|
| "Bootstrapping (general)": plot_td_backup,
|
| "n-step TD Backup": plot_nstep_td,
|
| "TD(lambda) & Eligibility Traces": plot_eligibility_traces,
|
| "SARSA Update": plot_sarsa_backup,
|
| "Q-Learning Update": plot_q_learning_backup,
|
| "Expected SARSA": plot_expected_sarsa_backup,
|
| "Double Q-Learning / Double DQN": plot_double_q,
|
| "Dueling DQN Architecture": plot_dueling_dqn,
|
| "Prioritized Experience Replay": plot_prioritized_replay,
|
| "Rainbow DQN Components": plot_rainbow_dqn,
|
| "Linear Function Approximation": plot_linear_fa,
|
| "Neural Network Layers (MLP, CNN, RNN, Transformer)": plot_nn_layers,
|
| "Computation Graph / Backpropagation Flow": plot_computation_graph,
|
| "Target Network": plot_target_network,
|
| "Policy Gradient Theorem": plot_policy_gradient_flow,
|
| "REINFORCE Update": plot_reinforce_flow,
|
| "Baseline / Advantage Subtraction": plot_advantage_scaled_grad,
|
| "Trust Region (TRPO)": plot_trpo_trust_region,
|
| "Proximal Policy Optimization (PPO)": plot_ppo_clip,
|
| "Actor-Critic Architecture": plot_actor_critic_arch,
|
| "Advantage Actor-Critic (A2C/A3C)": plot_a3c_multi_worker,
|
| "Soft Actor-Critic (SAC)": plot_sac_arch,
|
| "Twin Delayed DDPG (TD3)": plot_actor_critic_arch,
|
| "epsilon-Greedy Strategy": plot_epsilon_decay,
|
| "Softmax / Boltzmann Exploration": plot_softmax_exploration,
|
| "Upper Confidence Bound (UCB)": plot_ucb_confidence,
|
| "Intrinsic Motivation / Curiosity": plot_intrinsic_motivation,
|
| "Entropy Regularization": plot_entropy_bonus,
|
| "Options Framework": plot_options_framework,
|
| "Feudal Networks / Hierarchical Actor-Critic": plot_feudal_networks,
|
| "Skill Discovery": plot_skill_discovery,
|
| "Learned Dynamics Model": plot_world_model,
|
| "Model-Based Planning": plot_model_planning,
|
| "Imagination-Augmented Agents (I2A)": plot_imagination_rollout,
|
| "Offline Dataset": plot_offline_rl,
|
| "Conservative Q-Learning (CQL)": plot_cql_regularization,
|
| "Multi-Agent Interaction Graph": plot_multi_agent_interaction,
|
| "Centralized Training Decentralized Execution (CTDE)": plot_ctde,
|
| "Cooperative / Competitive Payoff Matrix": plot_payoff_matrix,
|
| "Reward Inference": plot_irl_reward_inference,
|
| "Generative Adversarial Imitation Learning (GAIL)": plot_gail_flow,
|
| "Meta-RL Architecture": plot_meta_rl_nested_loop,
|
| "Task Distribution Visualization": plot_task_distribution,
|
| "Experience Replay Buffer": plot_replay_buffer,
|
| "State Visitation / Occupancy Measure": plot_state_visitation,
|
| "Learning Curve": plot_learning_curve,
|
| "Regret / Cumulative Regret": plot_regret_curve,
|
| "Attention Mechanisms (Transformers in RL)": plot_attention_weights,
|
| "Diffusion Policy": plot_diffusion_policy,
|
| "Graph Neural Networks for RL": plot_gnn_rl,
|
| "World Model / Latent Space": plot_latent_space,
|
| "Convergence Analysis Plots": plot_convergence_log,
|
| "RL Algorithm Taxonomy": plot_rl_taxonomy_tree,
|
| "Probabilistic Graphical Model (RL as Inference)": plot_rl_as_inference_pgm,
|
| "Distributional RL (C51 / Categorical)": plot_distributional_rl_atoms,
|
| "Hindsight Experience Replay (HER)": plot_her_goal_relabeling,
|
| "Dyna-Q Architecture": plot_dyna_q_flow,
|
| "Noisy Networks (Parameter Noise)": plot_noisy_nets_parameters,
|
| "Intrinsic Curiosity Module (ICM)": plot_icm_curiosity,
|
| "V-trace (IMPALA)": plot_v_trace_impala,
|
| "QMIX Mixing Network": plot_qmix_mixing_net,
|
| "Saliency Maps / Attention on State": plot_saliency_heatmaps,
|
| "Action Selection Noise (OU vs Gaussian)": plot_action_selection_noise,
|
| "t-SNE / UMAP State Embeddings": plot_tsne_state_embeddings,
|
| "Loss Landscape Visualization": plot_loss_landscape,
|
| "Success Rate vs Steps": plot_success_rate_curve,
|
| "Hyperparameter Sensitivity Heatmap": plot_hyperparameter_sensitivity,
|
| "Action Persistence (Frame Skipping)": plot_action_persistence,
|
| "MuZero Dynamics Search Tree": plot_muzero_search_tree,
|
| "Policy Distillation": plot_policy_distillation,
|
| "Decision Transformer Token Sequence": plot_decision_transformer_tokens,
|
| "Performance Profiles (rliable)": plot_performance_profiles_rliable,
|
| "Safety Shielding / Barrier Functions": plot_safety_shielding,
|
| "Automated Curriculum Learning": plot_automated_curriculum,
|
| "Domain Randomization": plot_domain_randomization,
|
| "RL with Human Feedback (RLHF)": plot_rlhf_flow,
|
| "Successor Representations (SR)": plot_successor_representations,
|
| "Maximum Entropy IRL": plot_maxent_irl_trajectories,
|
| "Information Bottleneck": plot_information_bottleneck,
|
| "Evolutionary Strategies Population": plot_es_population_distribution,
|
| "Control Barrier Functions (CBF)": plot_cbf_safe_set,
|
| "Count-based Exploration Heatmap": plot_count_based_exploration,
|
| "Thompson Sampling Posteriors": plot_thompson_sampling,
|
| "Adversarial RL Interaction": plot_adversarial_rl_interaction,
|
| "Hierarchical Subgoal Trajectory": plot_hierarchical_subgoals,
|
| "Offline Action Distribution Shift": plot_offline_distribution_shift,
|
| "Random Network Distillation (RND)": plot_rnd_curiosity,
|
| "Batch-Constrained Q-learning (BCQ)": plot_bcq_offline_constraint,
|
| "Population-Based Training (PBT)": plot_pbt_evolution,
|
| "Recurrent State Flow (DRQN/R2D2)": plot_recurrent_state_flow,
|
| "Belief State in POMDPs": plot_belief_state_pomdp,
|
| "Multi-Objective Pareto Front": plot_pareto_front_morl,
|
| "Differential Value (Average Reward RL)": plot_differential_value_average_reward,
|
| "Distributed RL Cluster (Ray/RLLib)": plot_distributed_rl_cluster,
|
| "Neuroevolution Topology Evolution": plot_neuroevolution_topology,
|
| "Elastic Weight Consolidation (EWC)": plot_ewc_elastic_weights,
|
| "Successor Features (SF)": plot_successor_features,
|
| "Adversarial State Noise (Perception)": plot_adversarial_state_noise,
|
| "Behavioral Cloning (Imitation)": plot_behavioral_cloning_il,
|
| "Relational Graph State Representation": plot_relational_graph_state,
|
| "Quantum RL Circuit (PQC)": plot_quantum_rl_circuit,
|
| "Symbolic Policy Tree": plot_symbolic_expression_tree,
|
| "Differentiable Physics Gradient Flow": plot_differentiable_physics_gradient,
|
| "MARL Communication Channel": plot_marl_communication_channel,
|
| "Lagrangian Constraint Landscape": plot_lagrangian_multiplier_landscape,
|
| "MAXQ Task Hierarchy": plot_maxq_task_hierarchy,
|
| "ReAct Agentic Cycle": plot_react_cycle_thinking,
|
| "Synaptic Plasticity RL": plot_synaptic_plasticity_rl,
|
| "Guided Policy Search (GPS)": plot_guided_policy_search_gps,
|
| "Sim-to-Real Jitter & Latency": plot_sim2real_jitter_latency,
|
| "Deterministic Policy Gradient (DDPG) Flow": plot_ddpg_deterministic_gradient,
|
| "Dreamer Latent Imagination": plot_dreamer_latent_rollout,
|
| "UNREAL Auxiliary Tasks": plot_unreal_auxiliary_tasks,
|
| "Implicit Q-Learning (IQL) Expectile": plot_iql_expectile_loss,
|
| "Prioritized Sweeping": plot_prioritized_sweeping,
|
| "DAgger Expert Loop": plot_dagger_expert_loop,
|
| "Self-Predictive Representations (SPR)": plot_spr_self_prediction,
|
| "Joint Action Space": plot_joint_action_space,
|
| "Dec-POMDP Formal Model": plot_dec_pomdp_graph,
|
| "Bisimulation Metric": plot_bisimulation_metric,
|
| "Potential-Based Reward Shaping": plot_reward_shaping_phi,
|
| "Transfer RL: Source to Target": plot_transfer_rl_source_target,
|
| "Multi-Task Backbone Arch": plot_multi_task_backbone,
|
| "Contextual Bandit Pipeline": plot_contextual_bandit_pipeline,
|
| "Theoretical Regret Bounds": plot_regret_bounds_theoretical,
|
| "Soft Q Boltzmann Probabilities": plot_soft_q_heatmap,
|
| "Autonomous Driving RL Pipeline": plot_ad_rl_pipeline,
|
| "Policy action gradient comparison": plot_action_grad_comparison,
|
| "IRL: Feature Expectation Matching": plot_irl_feature_matching,
|
| "Apprenticeship Learning Loop": plot_apprenticeship_learning_loop,
|
| "Active Inference Loop": plot_active_inference_loop,
|
| "Bellman Residual Landscape": plot_bellman_residual_landscape,
|
| "Plan-to-Explore Uncertainty Map": plot_plan_to_explore_map,
|
| "Robust RL Uncertainty Set": plot_robust_rl_uncertainty_set,
|
| "HPO Bayesian Opt Cycle": plot_hpo_bayesian_opt_cycle,
|
| "Slate RL Recommendation": plot_slate_rl_reco_pipeline,
|
| "Fictitious Play Interaction": plot_game_theory_fictitious_play,
|
| "Universal RL Framework Diagram": plot_universal_rl_framework,
|
| "Offline Density Ratio Estimator": plot_offline_density_ratio,
|
| "Continual Task Interference Heatmap": plot_continual_task_interference,
|
| "Lyapunov Stability Safe Set": plot_lyapunov_safe_set,
|
| "Molecular RL (Atom Coordinates)": plot_molecular_rl_atoms,
|
| "MoE Multi-task Architecture": plot_moe_multi_task_arch,
|
| "CMA-ES Policy Search": plot_cma_es_distribution,
|
| "Elo Rating Preference Plot": plot_elo_rating_preference,
|
| "Explainable RL (SHAP Attribution)": plot_shap_lime_attribution,
|
| "PEARL Context Encoder": plot_pearl_context_encoder,
|
| "Medical RL Therapy Pipeline": plot_healthcare_rl_pipeline,
|
| "Supply Chain RL Pipeline": plot_supply_chain_rl,
|
| "Sim-to-Real SysID Loop": plot_sysid_safe_loop,
|
| "Transformer World Model": plot_transformer_world_model,
|
| "Network Traffic RL": plot_network_rl,
|
| "RLHF: PPO with Reference Policy": plot_rlhf_ppo_ref,
|
| "PSRO Meta-Game Update": plot_psro_meta_game,
|
| "DIAL: Differentiable Comm": plot_dial_comm_channel,
|
| "Fitted Q-Iteration Loop": plot_fqi_batch_loop,
|
| "CMDP Feasible Region": plot_cmdp_feasible_set,
|
| "MPC vs RL Planning": plot_mpc_vs_rl_horizon,
|
| "Learning to Optimize (L2O)": plot_l2o_meta_pipeline,
|
| "Smart Grid RL Management": plot_smart_grid_rl,
|
| "Quantum State Tomography RL": plot_quantum_tomography_rl,
|
| "Absolute Universal RL Pillar Map": plot_absolute_encyclopedia_map,
|
| "RL for Chip Placement": plot_chip_placement_rl,
|
| "RL Compiler Optimization (MLGO)": plot_compiler_mlgo,
|
| "RL for Theorem Proving": plot_theorem_proving_rl,
|
| "Diffusion-QL Offline RL": plot_diffusion_ql_loop,
|
| "Fairness-reward Pareto Frontier": plot_fairness_rl_pareto,
|
| "Differentially Private RL": plot_dp_rl_noise,
|
| "Smart Agriculture RL": plot_smart_agriculture_rl,
|
| "Climate Mitigation RL (Grid)": plot_climate_rl_grid,
|
| "AI Education (Knowledge Tracing)": plot_ai_education_tracing,
|
| "Decision SDE Flow": plot_decision_sde_flow,
|
| "Differentiable physics (Brax)": plot_diff_physics_brax,
|
| "Wireless Beamforming RL": plot_beamforming_rl,
|
| "Quantum Error Correction RL": plot_quantum_error_correction_rl,
|
| "Mean Field RL Interaction": plot_mean_field_rl,
|
| "Goal-GAN Curriculum": plot_goal_gan_hrl,
|
| "JEPA: Predictive Architecture": plot_jepa_arch,
|
| "CQL Value Penalty Landscape": plot_cql_penalty_surface,
|
| "Cybersecurity Attack-Defense RL": plot_cyber_attack_defense,
|
| "Causal Inverse RL Graph": plot_causal_irl,
|
| "VQE-RL Optimization": plot_vqe_rl,
|
| "De-novo Drug Discovery RL": plot_drug_discovery_rl,
|
| "Traffic Signal Coordination RL": plot_traffic_signal_coordination,
|
| "Mars Rover Pathfinding RL": plot_mars_rover_pathfinding,
|
| "Sports Player Movement RL": plot_sports_analytics_rl,
|
| "Cryptography Attack RL": plot_crypto_attack_rl,
|
| "Humanitarian Resource RL": plot_humanitarian_rl,
|
| "Video Compression RL (Rate-Distortion)": plot_video_compression_rl,
|
| "Kubernetes Auto-scaling RL": plot_kubernetes_scaling_rl,
|
| "Fluid Dynamics Flow Control RL": plot_fluid_dynamics_rl,
|
| "Structural Optimization RL": plot_structural_optimization_rl,
|
| "Human Decision Modeling (Prospect Theory)": plot_human_decision_rl,
|
| "Semantic Parsing RL": plot_semantic_parsing_rl,
|
| "Melody generation RL": plot_music_melody_rl,
|
| "Plasma Fusion Control RL": plot_plasma_control_rl,
|
| "Carbon Capture RL cycle": plot_carbon_capture_rl,
|
| "Swarm Robotics RL": plot_swarm_robotics_rl,
|
| "Legal Compliance RL Game": plot_legal_compliance_rl,
|
| "Physics-Informed RL (PINN)": plot_pinn_rl_loss,
|
| "Neuro-Symbolic RL": plot_neuro_symbolic_rl,
|
| "DeFi Liquidity Pool RL": plot_defi_liquidity_rl,
|
| "Dopamine Reward Prediction Error": plot_dopamine_rpe_curves,
|
| "Proprioceptive Sensory-Motor RL": plot_proprioceptive_rl_loop,
|
| "Augmented Reality Object Placement RL": plot_ar_placement_rl,
|
| "Sequential Bundle RL": plot_sequential_bundle_rl,
|
| "Online Gradient Descent vs RL": plot_ogd_vs_rl_gradient,
|
| "Active Learning: Query RL Selection": plot_active_learning_selection,
|
| "Federated RL global Aggregator": plot_federated_rl_tree,
|
| "Ultimate Universal RL Mastery Diagram": plot_ultimate_mastery_diagram
|
| }
|
|
|
| import sys
|
|
|
| for name, func in mapping.items():
|
|
|
| filename = re.sub(r'[^a-zA-Z0-9]', '_', name.lower()).strip('_')
|
| filename = re.sub(r'_+', '_', filename) + ".png"
|
| filepath = os.path.join(output_dir, filename)
|
|
|
| print(f"Generating: {filename} ...")
|
|
|
| plt.close('all')
|
|
|
| if func in [plot_reward_landscape, plot_loss_landscape]:
|
| fig = plt.figure(figsize=(10, 8))
|
| gs = GridSpec(1, 1, figure=fig)
|
| func(fig, gs)
|
| plt.savefig(filepath, bbox_inches='tight', dpi=100)
|
| plt.close(fig)
|
| continue
|
|
|
| fig, ax = plt.subplots(figsize=(10, 8), constrained_layout=True)
|
| func(ax)
|
| plt.savefig(filepath, bbox_inches='tight', dpi=100)
|
| plt.close(fig)
|
|
|
| print(f"\n[SUCCESS] Saved {len(mapping)} graphs to '{output_dir}/' directory.")
|
|
|
| if __name__ == "__main__":
|
| import sys
|
| if "--save" in sys.argv:
|
| save_all_graphs()
|
| else:
|
| main()
|
