text stringlengths 1 93.6k |
|---|
# <FILESEP>
|
import torch
|
import numpy as np
|
from scipy.spatial.distance import cdist
|
from sklearn.svm import SVC
|
from sklearn.neighbors import KNeighborsClassifier
|
from sklearn.metrics.pairwise import cosine_similarity
|
from sklearn.linear_model import LogisticRegression
|
def euclidean_dist(x, y):
|
# x: N x D
|
# y: M x D
|
n = x.size(0)
|
m = y.size(0)
|
d = x.size(1)
|
assert d == y.size(1)
|
x = x.unsqueeze(1).expand(n, m, d)
|
y = y.unsqueeze(0).expand(n, m, d)
|
return torch.pow(x - y, 2).sum(2)
|
def cosine_dist(x,y):
|
n = x.size(0)
|
m = y.size(0)
|
d = x.size(1)
|
assert d == y.size(1)
|
cosine_sim_list = []
|
for i in range(m):
|
y_tmp = y[i].unsqueeze(0)
|
x_tmp = x[0].unsqueeze(0)
|
#print(x_tmp.size(),y_tmp.size())
|
cosine_sim = torch.nn.functional.cosine_similarity(x_tmp,y_tmp)
|
cosine_sim_list.append(cosine_sim[0])
|
return torch.stack(cosine_sim_list)
|
def generate_prototypes_tensor_lowerdim(data):
|
'''
|
data: dict{'support_feature'[],'support_y'[],'query_feature'[],'query_y'[]}
|
return: prototype_ids, prototype_features
|
'''
|
support_feature, support_y, query_x, query_y = data['support_feature'], data['support_y'], data['query_feature'], data['query_y']
|
# get prototype ids and prototype_features
|
prototype_ids = []
|
prototype_features = []
|
dict = {}
|
for i in range(support_y.size()[0]):
|
classId = support_y[i]
|
video_feature = support_feature[i]
|
if classId not in dict.keys():
|
dict[classId] = [video_feature]
|
else:
|
dict[classId].append(video_feature)
|
for classId in dict.keys():
|
prototype_ids.append(classId)
|
prototype_feature = torch.stack(dict[classId])
|
prototype_feature = torch.mean(prototype_feature, axis=0)
|
prototype_features.append(prototype_feature)
|
prototype_features = torch.stack(prototype_features)
|
return (prototype_ids, prototype_features)
|
def one_shot_classifier_prototype_lowerdim(data):
|
'''
|
data: dict{'support_feature[],'support_y'[],'query_feature'[],'query_y'[]}
|
return : predicted_y
|
'''
|
# get input
|
support_feature, support_y, query_feature, query_y = data['support_feature'], data['support_y'], data['query_feature'], data['query_y']
|
# get prototypes_ids and prototype_features
|
prototype_ids, prototype_features = generate_prototypes_tensor_lowerdim(data)
|
# print(prototype_ids,prototype_features.shape)
|
# get distance
|
query_features = []
|
for i in range(query_y.size()[0]):
|
query_feature = query_feature[i]
|
# query_feature = np.mean(query_feature, axis=0)
|
query_features.append(query_feature)
|
query_features = torch.stack(query_features)
|
distance = euclidean_dist(query_features, prototype_features)
|
probability = torch.nn.functional.softmax(-distance)
|
import torch.nn as nn
|
criterion = nn.CrossEntropyLoss()
|
label= query_y.long()
|
loss = criterion(probability, label)
|
probability = probability.data.cpu().numpy()
|
predicted_y = np.argmax(probability, axis=1)
|
return predicted_y, loss
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.