-
Notifications
You must be signed in to change notification settings - Fork 96
/
naive_bayes.py
334 lines (310 loc) · 16.5 KB
/
naive_bayes.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
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
# -------------------------------------------------------------------------
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See License.txt in the project root for
# license information.
# --------------------------------------------------------------------------
import numpy as np
from ..proto import onnx_proto
from ..common._apply_operation import apply_add, apply_cast, apply_exp
from ..common._apply_operation import apply_log, apply_sub, apply_reshape
from ..common.data_types import Int64TensorType
from ..common._registration import register_converter
def convert_sklearn_naive_bayes(scope, operator, container):
# Computational graph:
#
# Note: In the following graph, variable names are in lower case
# characters only and operator names are in upper case characters.
# We borrow operator names from the official ONNX spec:
# https://github.com/onnx/onnx/blob/master/docs/Operators.md
# All variables are followed by their shape in [].
#
# Symbols:
# M: Number of instances
# N: Number of features
# C: Number of classes
# input(or x): input
# output(or y): output (There are two paths for producing output, one for
# string labels and the other one for int labels)
# output_probability: class probabilties
# feature_log_prob: Empirical log probability of features given a
# class, P(x_i|y)
# class_log_prior: Smoothed empirical log probability for each class
#
# Multinomial NB
# Equation:
# y = argmax (class_log_prior + X . feature_log_prob^T)
#
# Graph:
#
# input [M, N] -> MATMUL <- feature_log_prob.T [N, C]
# |
# V
# matmul_result [M, C] -> CAST <- onnx_proto.TensorProto.FLOAT
# |
# V
# cast_result [M, C] -> SUM <- class_log_prior [1, C]
# |
# .-----------------'
# |
# V
# sum_result [M, C] -> ARGMAX -> argmax_output [M, 1]
# |
# V
# classes [C] -----> ARRAYFEATUREEXTRACTOR
# |
# V (string labels)
# array_feature_extractor_result [M, 1] --------------------------.
# (int labels) | |
# V |
# CAST(to=onnx_proto.TensorProto.FLOAT) |
# | |
# V |
# cast2_result [M, 1] |
# | |
# V |
# output_shape [1] -> RESHAPE |
# | |
# V V
# reshaped_result [M,] .-----RESHAPE
# | |
# V V
# (to=onnx_proto.TensorProto.INT64)CAST --------> output [M,]
#
# Bernoulli NB
# Equation:
# y = argmax (class_log_prior + \sum neg_prob
# + X . (feature_log_prob - neg_prob))
# neg_prob = log( 1 - e ^ feature_log_prob)
#
# Graph:
#
# .---------------------------------------------------------.
# | |
# feature_log_prob.T [N, C] -> EXP -> exp_result [N, C] |
# | |
# .----------------------' |
# | |
# V V
# constant -> SUB -> sub_result [N, C] -> LOG -> neg_prob [N, C] -> SUB
# | |
# .---------' .---------'
# | |
# V V
# .----------- sum_neg_prob [1, C] <- REDUCE_SUM difference_matrix [N, C]
# | |
# | .----------------------------------'
# | |
# | V
# | input [M, N] -> MATMUL -> dot_product [M, C]
# | |
# | V
# '------------------------------------> SUM
# |
# V
# class_log_prior [1, C] -> SUM <- partial_sum_result [M, C]
# |
# V
# sum_result [M, C] -> ARGMAX -> argmax_output [M, 1]
# |
# V
# classes [C] -------> ARRAYFEATUREEXTRACTOR
# |
# .------------------------------------'
# |
# V (string labels)
# array_feature_extractor_result [M, 1] ----------------.
# (int labels) | |
# V |
# CAST(to=onnx_proto.TensorProto.FLOAT) |
# | |
# V |
# cast2_result [M, 1] |
# | |
# V |
# output_shape [1] -> RESHAPE |
# | |
# V V
# reshaped_result [M,] RESHAPE
# | |
# V |
# (to=onnx_proto.TensorProto.INT64)CAST -> output [M,] <-'
#
#
# If model's binarize attribute is not null, then input of
# Bernoulli NB is produced by the following graph:
#
# input [M, N] -> GREATER <- threshold [1]
# | |
# | V
# | condition [M, N] -> CAST(to=onnx_proto.TensorProto.FLOAT)
# | |
# | V
# | cast_values [M, N]
# | |
# V V
# CONSTANT_LIKE -> zero_tensor [M, N] -> ADD
# |
# V
# input [M, N] <- binarised_input [M, N]
#
# Sub-graph for probability calculation common to both Multinomial
# and Bernoulli Naive Bayes
#
# sum_result [M, C] -> REDUCELOGSUMEXP -> reduce_log_sum_exp_result [M,]
# | |
# | V
# | log_prob_shape [2] -> RESHAPE
# | |
# '------------> SUB <-- reshaped_log_prob [M, 1] <---'
# |
# V
# log_prob [M, C] -> EXP -> prob_tensor [M, C] -.
# |
# output_probability [M, C] <- ZIPMAP <---------------------'
nb = operator.raw_operator
class_log_prior = nb.class_log_prior_.astype('float32').reshape((1, -1))
feature_log_prob = nb.feature_log_prob_.T.astype('float32')
classes = nb.classes_
output_shape = (-1,)
class_log_prior_name = scope.get_unique_variable_name('class_log_prior')
feature_log_prob_name = scope.get_unique_variable_name('feature_log_prob')
sum_result_name = scope.get_unique_variable_name('sum_result')
argmax_output_name = scope.get_unique_variable_name('argmax_output')
cast2_result_name = scope.get_unique_variable_name('cast2_result')
reshaped_result_name = scope.get_unique_variable_name('reshaped_result')
classes_name = scope.get_unique_variable_name('classes')
reduce_log_sum_exp_result_name = scope.get_unique_variable_name(
'reduce_log_sum_exp_result')
log_prob_name = scope.get_unique_variable_name('log_prob')
array_feature_extractor_result_name = scope.get_unique_variable_name(
'array_feature_extractor_result')
class_type = onnx_proto.TensorProto.STRING
if np.issubdtype(nb.classes_.dtype, np.floating):
class_type = onnx_proto.TensorProto.INT32
classes = np.array(list(map(lambda x: int(x), classes)))
elif np.issubdtype(nb.classes_.dtype, np.signedinteger):
class_type = onnx_proto.TensorProto.INT32
else:
classes = np.array([s.encode('utf-8') for s in classes])
container.add_initializer(
feature_log_prob_name, onnx_proto.TensorProto.FLOAT,
feature_log_prob.shape, feature_log_prob.flatten())
container.add_initializer(classes_name, class_type, classes.shape, classes)
container.add_initializer(
class_log_prior_name, onnx_proto.TensorProto.FLOAT,
class_log_prior.shape, class_log_prior.flatten())
if container.target_opset < 6:
sum_op_version = 1
elif container.target_opset < 8:
sum_op_version = 6
else:
sum_op_version = 8
input_name = operator.inputs[0].full_name
if type(operator.inputs[0].type) == Int64TensorType:
cast_input_name = scope.get_unique_variable_name('cast_input')
apply_cast(scope, operator.input_full_names, cast_input_name,
container, to=onnx_proto.TensorProto.FLOAT)
input_name = cast_input_name
if operator.type == 'SklearnMultinomialNB':
matmul_result_name = scope.get_unique_variable_name('matmul_result')
container.add_node(
'MatMul', [input_name, feature_log_prob_name],
matmul_result_name, name=scope.get_unique_operator_name('MatMul'))
apply_add(scope, [matmul_result_name, class_log_prior_name],
sum_result_name, container, broadcast=1)
else:
constant_name = scope.get_unique_variable_name('constant')
exp_result_name = scope.get_unique_variable_name('exp_result')
sub_result_name = scope.get_unique_variable_name('sub_result')
neg_prob_name = scope.get_unique_variable_name('neg_prob')
sum_neg_prob_name = scope.get_unique_variable_name('sum_neg_prob')
difference_matrix_name = scope.get_unique_variable_name(
'difference_matrix')
dot_prod_name = scope.get_unique_variable_name('dot_prod')
partial_sum_result_name = scope.get_unique_variable_name(
'partial_sum_result')
container.add_initializer(constant_name, onnx_proto.TensorProto.FLOAT,
[], [1.0])
if nb.binarize is not None:
threshold_name = scope.get_unique_variable_name('threshold')
condition_name = scope.get_unique_variable_name('condition')
cast_values_name = scope.get_unique_variable_name('cast_values')
zero_tensor_name = scope.get_unique_variable_name('zero_tensor')
binarised_input_name = scope.get_unique_variable_name(
'binarised_input')
num_features = nb.feature_count_.shape[1]
M = operator.inputs[0].type.shape[0]
container.add_initializer(threshold_name,
onnx_proto.TensorProto.FLOAT,
[1], [nb.binarize])
container.add_initializer(
zero_tensor_name,
onnx_proto.TensorProto.FLOAT, [M, num_features],
np.zeros((M, num_features)).ravel())
container.add_node(
'Greater', [input_name, threshold_name],
condition_name, name=scope.get_unique_operator_name('Greater'),
op_version=9)
apply_cast(scope, condition_name, cast_values_name, container,
to=onnx_proto.TensorProto.FLOAT)
apply_add(scope, [zero_tensor_name, cast_values_name],
binarised_input_name, container, broadcast=1)
input_name = binarised_input_name
apply_exp(scope, feature_log_prob_name, exp_result_name, container)
apply_sub(scope, [constant_name, exp_result_name], sub_result_name,
container, broadcast=1)
apply_log(scope, sub_result_name, neg_prob_name, container)
container.add_node('ReduceSum', neg_prob_name,
sum_neg_prob_name, axes=[0],
name=scope.get_unique_operator_name('ReduceSum'))
apply_sub(scope, [feature_log_prob_name, neg_prob_name],
difference_matrix_name, container)
container.add_node(
'MatMul', [input_name, difference_matrix_name],
dot_prod_name, name=scope.get_unique_operator_name('MatMul'))
container.add_node(
'Sum', [sum_neg_prob_name, dot_prod_name],
partial_sum_result_name, op_version=sum_op_version,
name=scope.get_unique_operator_name('Sum'))
container.add_node(
'Sum', [partial_sum_result_name, class_log_prior_name],
sum_result_name, name=scope.get_unique_operator_name('Sum'),
op_version=sum_op_version)
container.add_node('ArgMax', sum_result_name,
argmax_output_name,
name=scope.get_unique_operator_name('ArgMax'), axis=1)
# Calculation of class probability
log_prob_shape = [-1, 1]
reshaped_log_prob_name = scope.get_unique_variable_name(
'reshaped_log_prob')
container.add_node('ReduceLogSumExp', sum_result_name,
reduce_log_sum_exp_result_name,
name=scope.get_unique_operator_name('ReduceLogSumExp'),
axes=[1], keepdims=0)
apply_reshape(scope, reduce_log_sum_exp_result_name,
reshaped_log_prob_name, container,
desired_shape=log_prob_shape)
apply_sub(scope, [sum_result_name, reshaped_log_prob_name], log_prob_name,
container, broadcast=1)
apply_exp(scope, log_prob_name, operator.outputs[1].full_name, container)
container.add_node(
'ArrayFeatureExtractor', [classes_name, argmax_output_name],
array_feature_extractor_result_name, op_domain='ai.onnx.ml',
name=scope.get_unique_operator_name('ArrayFeatureExtractor'))
# Reshape op does not seem to handle INT64 tensor even though it is
# listed as one of the supported types in the doc, so Cast was
# required here.
if class_type == onnx_proto.TensorProto.INT32:
apply_cast(scope, array_feature_extractor_result_name,
cast2_result_name, container,
to=onnx_proto.TensorProto.FLOAT)
apply_reshape(scope, cast2_result_name, reshaped_result_name,
container, desired_shape=output_shape)
apply_cast(scope, reshaped_result_name, operator.outputs[0].full_name,
container, to=onnx_proto.TensorProto.INT64)
else: # string labels
apply_reshape(scope, array_feature_extractor_result_name,
operator.outputs[0].full_name, container,
desired_shape=output_shape)
register_converter('SklearnMultinomialNB', convert_sklearn_naive_bayes)
register_converter('SklearnBernoulliNB', convert_sklearn_naive_bayes)