-
Notifications
You must be signed in to change notification settings - Fork 1
/
Experiment_FS.py
280 lines (225 loc) · 9.68 KB
/
Experiment_FS.py
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
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
from keras.datasets import mnist
from scipy.io import loadmat
#from utils import getSyntheticDataset
import os
from scipy.special import gamma,psi
from sklearn.neighbors import NearestNeighbors
import tensorflow as tf
from sklearn.cluster import KMeans
from scipy.spatial import distance_matrix
from scipy.optimize import linear_sum_assignment
from sklearn.neighbors import KNeighborsClassifier
from sklearn import preprocessing
from keras.models import Sequential
from keras.layers import Dense
from keras.utils.np_utils import to_categorical
from keras.callbacks import EarlyStopping
from scipy import ndimage
from scipy.linalg import det
from numpy import pi
import pickle
import scipy.io
from sklearn import svm
#from sklearn import cross_validation
from sklearn.metrics import accuracy_score
from sklearn.model_selection import KFold
def test_kmeans(x_test, y_test, faetures, number=20):
x = x_test.reshape(len(x_test), -1)
x = x[:,faetures]
n_clusters = len(set(y_test))
res = np.zeros(number)
for index in range(number):
km = KMeans(n_clusters=n_clusters, n_jobs=4, init='random')
y = km.fit_predict(x)
cost = np.zeros((n_clusters,n_clusters))
for i in range(n_clusters):
for j in range(n_clusters):
cost[i,j] = -sum(np.logical_and(y==i,y_test==j))
row_ind, col_ind = linear_sum_assignment(cost)
y_changed = 100*np.ones_like(y)
for i,j in zip(row_ind, col_ind):
y_changed[y==i] = j
res[index] = float(np.sum(y_changed==y_test))*100./float(len(y))
return (np.mean(res), np.std(res))
def test_knn(x_train, y_train, x_test, y_test, features):
x_test = x_test.reshape(len(x_test), -1)
x_train = x_train.reshape(len(x_train), -1)
x_test = x_test[:,features]
x_train = x_train[:,features]
knn = KNeighborsClassifier(n_neighbors=1, n_jobs=4)
knn.fit(x_train, y_train)
y = knn.predict(x_test)
return np.sum(y==y_test)*100/len(y)
def test_softmax(x_train, y_train, x_test, y_test, features):
early_stopping = EarlyStopping(patience=2)
x_test = x_test.reshape(len(x_test), -1)
x_train = x_train.reshape(len(x_train), -1)
x_test = x_test[:,features]
x_train = x_train[:,features]
num_classes = len(np.unique(y_train))
model = Sequential([Dense(num_classes, input_dim = len(features), activation='softmax')])
model.compile(optimizer='adam',loss = 'categorical_crossentropy', metrics=['accuracy'])
y_binary_train = to_categorical(y_train, num_classes=num_classes)
y_binary_test = to_categorical(y_test, num_classes=num_classes)
model.fit(x_train, y_binary_train, epochs=1000, validation_split=.15, verbose=0)
result = model.evaluate(x_test, y_binary_test)
return result[1]
def nearest_distances(X, k=1):
'''
X = array(N,M)
N = number of points
M = number of dimensions
returns the distance to the kth nearest neighbor for every point in X
'''
knn = NearestNeighbors(n_neighbors=k)
knn.fit(X)
d, _ = knn.kneighbors(X) # the first nearest neighbor is itself
return d[:, -1] # returns the distance to the kth nearest neighbor
def mutual_information(variables, k=1):
'''
Returns the mutual information between any number of variables.
Each variable is a matrix X = array(n_samples, n_features)
where
n = number of samples
dx,dy = number of dimensions
Optionally, the following keyword argument can be specified:
k = number of nearest neighbors for density estimation
Example: mutual_information((X, Y)), mutual_information((X, Y, Z), k=5)
'''
if len(variables) < 2:
raise AttributeError(
"Mutual information must involve at least 2 variables")
all_vars = np.hstack(variables)
return (sum([entropy(X, k=k) for X in variables])
- entropy(all_vars, k=k))
def entropy(X, k=1):
''' Returns the entropy of the X.
Parameters
===========
X : array-like, shape (n_samples, n_features)
The data the entropy of which is computed
k : int, optional
number of nearest neighbors for density estimation
Notes
======
Kozachenko, L. F. & Leonenko, N. N. 1987 Sample estimate of entropy
of a random vector. Probl. Inf. Transm. 23, 95-101.
See also: Evans, D. 2008 A computationally efficient estimator for
mutual information, Proc. R. Soc. A 464 (2093), 1203-1215.
and:
Kraskov A, Stogbauer H, Grassberger P. (2004). Estimating mutual
information. Phys Rev E 69(6 Pt 2):066138.
'''
# Distance to kth nearest neighbor
r = nearest_distances(X, k) # squared distances
n, d = X.shape
volume_unit_ball = (pi**(.5*d)) / gamma(.5*d + 1)
'''
F. Perez-Cruz, (2008). Estimation of Information Theoretic Measures
for Continuous Random Variables. Advances in Neural Information
Processing Systems 21 (NIPS). Vancouver (Canada), December.
return d*mean(log(r))+log(volume_unit_ball)+log(n-1)-log(k)
'''
return (d*np.mean(np.log(r + np.finfo(X.dtype).eps))
+ np.log(volume_unit_ball) + psi(n) - psi(k))
def maxrel_minred(x_train , weights , N , Bestfea_nums):
# N select the N wieghts with max values
ind1 = np.argpartition(weights, -N)[-N:]
selectedN = x_train[:,ind1]
MI_scores = np.zeros(N)
for i in range(N):
X1 = selectedN[:,i]
X2 = np.delete(selectedN, i , 1)
X1.reshape(-1,1)
X_z = np.zeros(X1.shape[0])
h = np.column_stack((X1,X_z))
d1 = nearest_distances(h , k = 15)
d2 = nearest_distances(X2 , k = 15)
MI_scores[i] = (np.sum(np.log(d1))+(99)*np.sum(np.log(d2)))/180
ind2 = np.argpartition(MI_scores, -1)[-1:]
bestfea_ind = ind2
alpha = 0.5
phi = np.array([]) # max information - min redundancy score
mi = np.array([])
for i in range(Bestfea_nums-1):
phi = np.array([])
for j in range(N):
if j in bestfea_ind:
phi = np.append(phi , - float('inf'))
else:
X1 = selectedN[:,j]
#dummy_ind = np.append(bestfea_ind, j)
X2 = selectedN[:,bestfea_ind]
X_z = np.zeros(X1.shape[0])
# concat a zero vector with data sets
h1 = np.column_stack((X1,X_z))
h2 = np.column_stack((X2,X_z))
d1 = nearest_distances(h1 , k = 15)
d2 = nearest_distances(h2 , k = 15)
miscor = (np.sum(np.log(d1))+(len(bestfea_ind)-1)*np.sum(np.log(d2)))/180
mi = np.append(mi,miscor)
phi = np.append(phi , (MI_scores[j] + alpha*mi[-1]))
new_ind = np.argpartition(phi, -1)[-1:]
bestfea_ind = np.append(bestfea_ind,new_ind)
finalbest_ind_mi = ind1[(bestfea_ind)]
finalbest_ind_aefs = np.argpartition(weights, -Bestfea_nums)[-Bestfea_nums:]
return finalbest_ind_mi , finalbest_ind_aefs
if __name__ == '__main__':
alpha = 0.001
beta = 0.01
dataset = 'face'
if(dataset == 'face'):
data = loadmat('E:/MSc/TensorFlow Learn/AEFS/warpPIE10P.mat')
X = data['X']/255.
#x_train = X[0:180,]
#x_test = X[180:209,]
Y = data['Y']-1
#y_train = Y[0:180]
#y_test = Y[180:209]
n_splits = 10
kf = KFold(n_splits)
input_shape = (44,55)
input_dim = 44*55
hidden = 128
# load weights
weights = np.load('E:/MSc/TensorFlow Learn/AEFS/weights.npy')
N = 100
Bestfea_nums = 40
acc_mi = np.zeros(n_splits)
acc_aefs = np.zeros(n_splits)
counter = 0
for train_index, test_index in kf.split(X):
#print("TRAIN:", train_index, "TEST:", test_index)
x_train, x_test = X[train_index], X[test_index]
y_train, y_test = Y[train_index], Y[test_index]
[fea_idx_mi , fea_idx_aefs] = maxrel_minred(x_train , weights , N , Bestfea_nums)
mi_mask =np.zeros(input_dim)
mi_mask[fea_idx_mi] = 1
bestfea_mi_image = np.multiply(mi_mask, weights)
aefs_mask = np.zeros(input_dim)
aefs_mask[fea_idx_aefs]=1
bestfea_aefs_image = np.multiply(aefs_mask , weights)
fig = plt.figure(figsize=(10,3))
ax = fig.add_subplot(131)
weights_martrix = weights.reshape(44,55)
ax.imshow(weights_martrix)
ax = fig.add_subplot(132)
ax.imshow(bestfea_mi_image.reshape(44,55))
ax = fig.add_subplot(133)
ax.imshow(bestfea_aefs_image.reshape(44,55))
## evalute the feature selection results of AEFS + MI
clf_mi = svm.LinearSVC() # linear SVM
clf_aefs = svm.LinearSVC()
clf_mi.fit(x_train[:, fea_idx_mi ], y_train)
clf_aefs.fit(x_train[:, fea_idx_aefs ], y_train)
y_predict_mi = clf_mi.predict(x_test[:,fea_idx_mi])
y_predict_aefs = clf_aefs.predict(x_test[:,fea_idx_aefs])
# obtain the classification accuracy on the test data
acc_mi[counter] = accuracy_score(y_test, y_predict_mi)
acc_aefs[counter] = accuracy_score(y_test , y_predict_aefs)
counter = counter+1
#print('mi acc' , acc_mi)
#print('aefs_acc' , acc_aefs)