-
Notifications
You must be signed in to change notification settings - Fork 0
/
classifiers.py
270 lines (206 loc) · 8.57 KB
/
classifiers.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
'''
Created on Sep 10, 2015
@author: Mustafa
'''
from abc import abstractmethod
from scipy.sparse import diags, issparse
from sklearn.linear_model import LogisticRegression, Lasso, Ridge, LinearRegression
from sklearn.naive_bayes import MultinomialNB, BernoulliNB, GaussianNB
from sklearn.svm import LinearSVC, LinearSVR
from sklearn.utils.validation import check_array
import numpy as np
class TransparentModel(object):
@abstractmethod
def predict_evidences(self, X):
"""Compute evidences for class 0 and class 1 and return them"""
@abstractmethod
def get_weights(self):
"""Return the weights from class 1's perspective.
Negative weight contributes to class 0,
positive weight contributes to class 1"""
@abstractmethod
def get_bias(self):
"""Return the class bias, if any, from class 1's perspective.
Negative value contributes to class 0,
positive value contributes to class 1"""
def compute_evidences_nonnegative_matrix(weights, X):
X = check_array(X, accept_sparse="csr")
neg_weights = weights * (weights < 0)
pos_weights = weights * (weights > 0)
if issparse(X):
neg_evi = X * neg_weights
pos_evi = X * pos_weights
else:
neg_evi = np.dot(X, neg_weights)
pos_evi = np.dot(X, pos_weights)
return neg_evi, pos_evi
def compute_evidences(weights, X):
X = check_array(X, accept_sparse="csr")
weights_diags = diags(weights, 0)
dm = X * weights_diags
if issparse(dm):
pos_evi = dm.multiply(dm > 0).sum(1).A1
neg_evi = dm.multiply(dm < 0).sum(1).A1
else:
pos_evi = np.multiply(dm, dm > 0).sum(1)
neg_evi = np.multiply(dm, dm < 0).sum(1)
return neg_evi, pos_evi
##############
# REGRESSION #
##############
class TransparentLasso(TransparentModel, Lasso):
'''
Transparent Lasso
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_, X)
else:
return compute_evidences(self.coef_, X)
def get_weights(self):
return self.coef_
def get_bias(self):
return self.intercept_
class TransparentRidge(TransparentModel, Ridge):
'''
Transparent Ridge regression
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_, X)
else:
return compute_evidences(self.coef_, X)
def get_weights(self):
return self.coef_
def get_bias(self):
return self.intercept_
class TransparentLinearRegression(TransparentModel, LinearRegression):
'''
Transparent linear regression
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_, X)
else:
return compute_evidences(self.coef_, X)
def get_weights(self):
return self.coef_
def get_bias(self):
return self.intercept_
class TransparentLinearSVR(TransparentModel, LinearSVR):
'''
Transparent linear SVR
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_, X)
else:
return compute_evidences(self.coef_, X)
def get_weights(self):
return self.coef_
def get_bias(self):
return self.intercept_
##################
# CLASSIFICATION #
##################
class TransparentLogisticRegression(TransparentModel, LogisticRegression):
'''
Transparent logistic regression
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_[0], X)
else:
return compute_evidences(self.coef_[0], X)
def get_weights(self):
return self.coef_[0]
def get_bias(self):
return self.intercept_[0]
class TransparentLinearSVC(TransparentModel, LinearSVC):
'''
Transparent Linear SVC
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() > 0:
return compute_evidences_nonnegative_matrix(self.coef_[0], X)
else:
return compute_evidences(self.coef_[0], X)
def get_weights(self):
return self.coef_[0]
def get_bias(self):
return self.intercept_[0]
class TransparentMultinomialNB(TransparentModel, MultinomialNB):
'''
Transparent multinomial naive Bayes
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse="csr")
if X.min() < 0:
raise ValueError("Multinomial naive Bayes cannot be used with negative feature values.")
weights = self.feature_log_prob_[1] - self.feature_log_prob_[0]
return compute_evidences_nonnegative_matrix(weights, X)
def get_weights(self):
return self.feature_log_prob_[1] - self.feature_log_prob_[0]
def get_bias(self):
return self.class_log_prior_[1] - self.class_log_prior_[0]
class TransparentBernoulliNB(TransparentModel, BernoulliNB):
'''
Transparent Bernoulli naive Bayes
'''
def predict_evidences(self, X):
p_neg_evi, p_pos_evi, a_neg_evi, a_pos_evi = self.predict_presence_absence_evidences(X)
return p_neg_evi + a_neg_evi, p_pos_evi + a_pos_evi
def predict_presence_absence_evidences(self, X):
X = check_array(X, accept_sparse="csr")
absence_log_prob_ = np.log(1 - np.exp(self.feature_log_prob_))
presence_log_ratios = self.feature_log_prob_[1] - self.feature_log_prob_[0]
absence_log_ratios = absence_log_prob_[1] - absence_log_prob_[0]
presence_neg_log_ratios = presence_log_ratios * (presence_log_ratios<0)
presence_pos_log_ratios = presence_log_ratios * (presence_log_ratios>0)
if issparse(X):
p_neg_evi = X * presence_neg_log_ratios
p_pos_evi = X * presence_pos_log_ratios
else:
p_neg_evi = np.dot(X, presence_neg_log_ratios)
p_pos_evi = np.dot(X, presence_pos_log_ratios)
absence_neg_log_ratios = absence_log_ratios * (absence_log_ratios<0)
absence_pos_log_ratios = absence_log_ratios * (absence_log_ratios>0)
default_a_neg_evi = absence_neg_log_ratios.sum()
default_a_pos_evi = absence_pos_log_ratios.sum()
if issparse(X):
a_neg_evi = -(X * absence_neg_log_ratios) + default_a_neg_evi
a_pos_evi = -(X * absence_pos_log_ratios) + default_a_pos_evi
else:
a_neg_evi = -np.dot(X, absence_neg_log_ratios) + default_a_neg_evi
a_pos_evi = -np.dot(X, absence_pos_log_ratios) + default_a_pos_evi
return p_neg_evi, p_pos_evi, a_neg_evi, a_pos_evi
def get_bias(self):
return self.class_log_prior_[1] - self.class_log_prior_[0]
def get_weights(self, presence=True):
if presence:
return self.feature_log_prob_[1] - self.feature_log_prob_[0]
else:
absence_log_prob_ = np.log(1 - np.exp(self.feature_log_prob_))
return absence_log_prob_[1] - absence_log_prob_[0]
class TransparentGaussianNB(TransparentModel, GaussianNB):
'''
Transparent Gaussian naive Bayes
'''
def predict_evidences(self, X):
X = check_array(X, accept_sparse=None) # For now, do not accept sparse
evi1 = -0.5*np.log(self.sigma_[1, :])
evi1 += 0.5*np.log(self.sigma_[0, :])
evi2 = -((X - self.theta_[1, :]) ** 2)/ (2*self.sigma_[1, :])
evi2 += ((X - self.theta_[0, :]) ** 2)/ (2*self.sigma_[0, :])
evi = evi1 + evi2
pos_evi = np.sum(evi*(evi > 0), 1)
neg_evi = np.sum(evi*(evi < 0), 1)
return neg_evi, pos_evi
def get_bias(self):
return np.log(self.class_prior_[1]) - np.log(self.class_prior_[0])