File size: 14,758 Bytes
b8d143b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bfbdc0b
b8d143b
 
 
 
 
 
 
 
 
 
bfbdc0b
 
 
 
 
 
 
b8d143b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5ac76c0
 
 
 
 
b8d143b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9b6c85d
 
b8d143b
 
 
 
 
 
9b6c85d
b8d143b
 
 
 
 
bb33480
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9b6c85d
 
 
 
bb33480
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b8d143b
bb33480
 
b8d143b
bb33480
 
 
b8d143b
 
 
 
 
 
 
 
bb33480
b8d143b
 
 
 
 
 
 
bb33480
b8d143b
 
 
 
bb33480
b8d143b
 
 
bb33480
 
 
 
 
 
 
 
 
b8d143b
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
"""
Deep learning classifier comparison for AURIS.

Trains and evaluates multiple neural network architectures on
the 47 extracted audio features using stratified k-fold CV.

Architectures:
  1. Deep MLP (512-256-128-64) with BatchNorm + Dropout
  2. 1D-CNN on feature vector (treats features as 1D signal)
  3. Residual MLP (skip connections)
  4. Attention MLP (self-attention over feature groups)

Usage:
    python -m app.training.train_deep_classifiers ../DataSet/features.csv
"""

from __future__ import annotations

import csv
import json
import sys
import time
from pathlib import Path

import numpy as np
import torch
import torch.nn as nn
from torch.utils.data import DataLoader, TensorDataset
from sklearn.model_selection import StratifiedKFold
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import accuracy_score, roc_auc_score, f1_score, precision_score, recall_score, roc_curve

DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
SEED = 42
N_FOLDS = 5
EPOCHS = 100
PATIENCE = 10
BATCH_SIZE = 64
LR = 1e-3


def _optimal_threshold(y_true: np.ndarray, y_prob: np.ndarray) -> float:
    """Youden's J: threshold maximising sensitivity + specificity - 1."""
    fpr, tpr, thresholds = roc_curve(y_true, y_prob)
    j = tpr - fpr
    return float(thresholds[np.argmax(j)])


def set_seed(seed: int = SEED) -> None:
    np.random.seed(seed)
    torch.manual_seed(seed)
    if torch.cuda.is_available():
        torch.cuda.manual_seed_all(seed)


class DeepMLP(nn.Module):
    def __init__(self, n_features: int) -> None:
        super().__init__()
        self.net = nn.Sequential(
            nn.Linear(n_features, 512),
            nn.BatchNorm1d(512),
            nn.ReLU(),
            nn.Dropout(0.4),
            nn.Linear(512, 256),
            nn.BatchNorm1d(256),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(256, 128),
            nn.BatchNorm1d(128),
            nn.ReLU(),
            nn.Dropout(0.2),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 1),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.net(x).squeeze(-1)


class Conv1DClassifier(nn.Module):
    def __init__(self, n_features: int) -> None:
        super().__init__()
        self.conv = nn.Sequential(
            nn.Conv1d(1, 64, kernel_size=5, padding=2),
            nn.BatchNorm1d(64),
            nn.ReLU(),
            nn.Conv1d(64, 128, kernel_size=3, padding=1),
            nn.BatchNorm1d(128),
            nn.ReLU(),
            nn.AdaptiveAvgPool1d(1),
        )
        self.fc = nn.Sequential(
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(64, 1),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = x.unsqueeze(1)  # (B, 1, F)
        x = self.conv(x).squeeze(-1)  # (B, 128)
        return self.fc(x).squeeze(-1)


class ResidualBlock(nn.Module):
    def __init__(self, dim: int, dropout: float = 0.2) -> None:
        super().__init__()
        self.block = nn.Sequential(
            nn.Linear(dim, dim),
            nn.BatchNorm1d(dim),
            nn.ReLU(),
            nn.Dropout(dropout),
            nn.Linear(dim, dim),
            nn.BatchNorm1d(dim),
        )
        self.relu = nn.ReLU()

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.relu(x + self.block(x))


class ResidualMLP(nn.Module):
    def __init__(self, n_features: int) -> None:
        super().__init__()
        self.input_proj = nn.Sequential(
            nn.Linear(n_features, 256),
            nn.BatchNorm1d(256),
            nn.ReLU(),
        )
        self.res_blocks = nn.Sequential(
            ResidualBlock(256, 0.3),
            ResidualBlock(256, 0.2),
            ResidualBlock(256, 0.1),
        )
        self.head = nn.Sequential(
            nn.Linear(256, 64),
            nn.ReLU(),
            nn.Linear(64, 1),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.input_proj(x)
        x = self.res_blocks(x)
        return self.head(x).squeeze(-1)


class AttentionMLP(nn.Module):
    def __init__(self, n_features: int) -> None:
        super().__init__()
        self.proj = nn.Linear(n_features, 256)
        self.attn = nn.MultiheadAttention(256, num_heads=4, batch_first=True)
        self.norm = nn.LayerNorm(256)
        self.head = nn.Sequential(
            nn.Linear(256, 128),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(128, 1),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.proj(x)
        x = x.unsqueeze(1)  # (B, 1, 256)
        x_chunk = x.expand(-1, 4, -1)  # (B, 4, 256) - create sequence
        attn_out, _ = self.attn(x_chunk, x_chunk, x_chunk)
        x = self.norm(attn_out.mean(dim=1))  # (B, 256)
        return self.head(x).squeeze(-1)


def load_data(csv_path: str | Path) -> tuple[np.ndarray, np.ndarray, list[str]]:
    _EXCLUDE = {"file_path", "label_int", "duration_sec", "sample_rate"}
    rows, labels = [], []
    with open(csv_path, "r", encoding="utf-8") as f:
        reader = csv.DictReader(f)
        feature_cols = [c for c in reader.fieldnames if c not in _EXCLUDE]
        for row in reader:
            vals = []
            for col in feature_cols:
                try:
                    vals.append(float(row[col]))
                except (ValueError, KeyError):
                    vals.append(0.0)
            rows.append(vals)
            labels.append(int(row["label_int"]))
    X = np.nan_to_num(np.array(rows, dtype=np.float32), nan=0.0)
    y = np.array(labels, dtype=np.int32)
    return X, y, feature_cols


def train_one_fold(
    model: nn.Module,
    X_train: np.ndarray, y_train: np.ndarray,
    X_val: np.ndarray, y_val: np.ndarray,
) -> tuple[float, np.ndarray]:
    scaler = StandardScaler()
    X_tr = scaler.fit_transform(X_train)
    X_v = scaler.transform(X_val)

    train_ds = TensorDataset(
        torch.tensor(X_tr, dtype=torch.float32),
        torch.tensor(y_train, dtype=torch.float32),
    )
    val_X = torch.tensor(X_v, dtype=torch.float32).to(DEVICE)
    val_y = torch.tensor(y_val, dtype=torch.float32)

    loader = DataLoader(train_ds, batch_size=BATCH_SIZE, shuffle=True)
    model = model.to(DEVICE)
    optimizer = torch.optim.AdamW(model.parameters(), lr=LR, weight_decay=1e-4)
    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
        optimizer, mode="max", factor=0.5, patience=5
    )
    # pos_weight compensates for class imbalance (n_neg / n_pos)
    n_pos = max(int(y_train.sum()), 1)
    n_neg = len(y_train) - n_pos
    pos_weight = torch.tensor([n_neg / n_pos], dtype=torch.float32).to(DEVICE)
    criterion = nn.BCEWithLogitsLoss(pos_weight=pos_weight)

    best_auc = 0.0
    best_probs = None
    patience_ctr = 0

    for epoch in range(EPOCHS):
        model.train()
        for bx, by in loader:
            bx, by = bx.to(DEVICE), by.to(DEVICE)
            optimizer.zero_grad()
            logits = model(bx)
            loss = criterion(logits, by)
            loss.backward()
            optimizer.step()

        model.eval()
        with torch.no_grad():
            v_logits = model(val_X)
            v_probs = torch.sigmoid(v_logits).cpu().numpy()

        auc = roc_auc_score(y_val, v_probs)
        scheduler.step(auc)

        if auc > best_auc:
            best_auc = auc
            best_probs = v_probs.copy()
            patience_ctr = 0
        else:
            patience_ctr += 1
            if patience_ctr >= PATIENCE:
                break

    return best_auc, best_probs


def evaluate_cv(
    model_class: type,
    X: np.ndarray, y: np.ndarray,
    n_features: int,
) -> dict:
    cv = StratifiedKFold(n_splits=N_FOLDS, shuffle=True, random_state=SEED)
    all_probs = np.zeros(len(y))
    aucs = []
    t0 = time.time()

    for fold, (train_idx, val_idx) in enumerate(cv.split(X, y)):
        set_seed(SEED + fold)
        model = model_class(n_features)
        auc, probs = train_one_fold(
            model,
            X[train_idx], y[train_idx],
            X[val_idx], y[val_idx],
        )
        all_probs[val_idx] = probs
        aucs.append(auc)
        print(f"    Fold {fold+1}: AUC={auc:.4f}")

    elapsed = time.time() - t0
    threshold = _optimal_threshold(y, all_probs)
    y_pred = (all_probs >= threshold).astype(int)
    return {
        "accuracy": round(float(accuracy_score(y, y_pred)), 4),
        "precision": round(float(precision_score(y, y_pred, zero_division=0)), 4),
        "recall": round(float(recall_score(y, y_pred, zero_division=0)), 4),
        "f1": round(float(f1_score(y, y_pred, zero_division=0)), 4),
        "roc_auc": round(float(roc_auc_score(y, all_probs)), 4),
        "optimal_threshold": round(threshold, 4),
        "fold_aucs": [round(a, 4) for a in aucs],
        "train_time_sec": round(elapsed, 1),
    }


class TorchSklearnWrapper:
    """
    Sklearn-compatible wrapper for trained PyTorch classifiers.
    Saves model class name + state dict so it can be pickled and reloaded.
    """

    def __init__(
        self,
        model_class: type,
        n_features: int,
        state_dict: dict,
        scaler: StandardScaler,
    ) -> None:
        self.model_class_name = model_class.__name__
        self._model_class = model_class
        self.n_features = n_features
        self.state_dict = state_dict
        self.scaler = scaler
        self.n_features_in_ = n_features

    def _build_model(self) -> nn.Module:
        model = self._model_class(self.n_features)
        model.load_state_dict(self.state_dict)
        model.eval()
        return model

    def predict_proba(self, X: np.ndarray) -> np.ndarray:
        model = self._build_model().to("cpu")
        X_scaled = self.scaler.transform(X)
        x_t = torch.tensor(X_scaled, dtype=torch.float32)
        with torch.no_grad():
            logits = model(x_t)
            probs = torch.sigmoid(logits).numpy().flatten()
        return np.column_stack([1.0 - probs, probs])

    def __getstate__(self) -> dict:
        state = self.__dict__.copy()
        state.pop("_model_class", None)
        return state

    def __setstate__(self, state: dict) -> None:
        self.__dict__.update(state)
        # Re-attach class from global lookup
        _CLASS_MAP = {
            "DeepMLP": DeepMLP,
            "Conv1DClassifier": Conv1DClassifier,
            "ResidualMLP": ResidualMLP,
            "AttentionMLP": AttentionMLP,
        }
        self._model_class = _CLASS_MAP.get(self.model_class_name, DeepMLP)


def train_final_model(
    model_class: type,
    X: np.ndarray,
    y: np.ndarray,
    epochs: int = EPOCHS,
    patience: int = PATIENCE,
) -> TorchSklearnWrapper:
    """Train model on full dataset and return sklearn-compatible wrapper."""
    from sklearn.model_selection import train_test_split

    scaler = StandardScaler()
    X_tr_raw, X_val_raw, y_tr, y_val = train_test_split(
        X, y, test_size=0.1, stratify=y, random_state=SEED
    )
    X_tr = scaler.fit_transform(X_tr_raw)
    X_v = scaler.transform(X_val_raw)

    n_features = X.shape[1]
    model = model_class(n_features).to(DEVICE)
    optimizer = torch.optim.AdamW(model.parameters(), lr=LR, weight_decay=1e-4)
    n_pos = max(int(y_tr.sum()), 1)
    n_neg = len(y_tr) - n_pos
    pos_weight = torch.tensor([n_neg / n_pos], dtype=torch.float32).to(DEVICE)
    criterion = nn.BCEWithLogitsLoss(pos_weight=pos_weight)
    loader = DataLoader(
        TensorDataset(
            torch.tensor(X_tr, dtype=torch.float32),
            torch.tensor(y_tr, dtype=torch.float32),
        ),
        batch_size=BATCH_SIZE,
        shuffle=True,
    )
    val_X = torch.tensor(X_v, dtype=torch.float32).to(DEVICE)
    val_y = torch.tensor(y_val, dtype=torch.float32)

    best_auc = 0.0
    best_state = None
    patience_ctr = 0

    for epoch in range(epochs):
        model.train()
        for bx, by in loader:
            bx, by = bx.to(DEVICE), by.to(DEVICE)
            optimizer.zero_grad()
            criterion(model(bx), by).backward()
            optimizer.step()

        model.eval()
        with torch.no_grad():
            probs = torch.sigmoid(model(val_X)).cpu().numpy()
        auc = roc_auc_score(val_y.numpy(), probs)
        if auc > best_auc:
            best_auc = auc
            best_state = {k: v.cpu().clone() for k, v in model.state_dict().items()}
            patience_ctr = 0
        else:
            patience_ctr += 1
            if patience_ctr >= patience:
                break

    return TorchSklearnWrapper(model_class, n_features, best_state or model.state_dict(), scaler)


def _safe_name(name: str) -> str:
    return name.lower().replace(" ", "_").replace("(", "").replace(")", "").replace("-", "_")


def main() -> None:
    import pickle

    csv_path = sys.argv[1] if len(sys.argv) > 1 else "../DataSet/features.csv"
    out_dir = Path(sys.argv[2]) if len(sys.argv) > 2 else Path("models")
    out_dir.mkdir(parents=True, exist_ok=True)

    print(f"Device: {DEVICE}")
    print(f"Loading: {csv_path}")

    X, y, feature_cols = load_data(csv_path)
    n_features = X.shape[1]
    print(f"Samples: {len(y)}, Features: {n_features}")
    print(f"AI: {np.sum(y == 1)}, Human: {np.sum(y == 0)}")

    model_classes = {
        "Deep MLP (512-256-128-64)": DeepMLP,
        "1D-CNN": Conv1DClassifier,
        "Residual MLP (3 blocks)": ResidualMLP,
        "Attention MLP": AttentionMLP,
    }

    all_results = {}
    for name, cls in model_classes.items():
        print(f"\n{'='*60}")
        print(f"  {name}")
        print(f"{'='*60}")
        result = evaluate_cv(cls, X, y, n_features)
        all_results[name] = {**result, "type": "deep_learning"}
        print(f"  => Acc={result['accuracy']:.4f}  AUC={result['roc_auc']:.4f}  "
              f"F1={result['f1']:.4f}  Time={result['train_time_sec']:.0f}s")

        print(f"  Training final model for {name}...")
        wrapper = train_final_model(cls, X, y)
        pkl_path = out_dir / f"model_dl_{_safe_name(name)}.pkl"
        with open(pkl_path, "wb") as f:
            pickle.dump(wrapper, f)
        all_results[name]["model_path"] = str(pkl_path)
        print(f"  Saved: {pkl_path}")

    out_path = out_dir / "deep_learning_results.json"
    with open(out_path, "w") as f:
        json.dump(all_results, f, indent=2)
    print(f"\nResults saved: {out_path}")

    print(f"\n{'='*60}")
    print("  SUMMARY")
    print(f"{'='*60}")
    for name, r in sorted(all_results.items(), key=lambda x: -x[1]["roc_auc"]):
        print(f"  {name:35s} AUC={r['roc_auc']:.4f}  Acc={r['accuracy']:.4f}")


if __name__ == "__main__":
    main()