-
Notifications
You must be signed in to change notification settings - Fork 0
/
my_model_selectors.py
182 lines (154 loc) · 7.73 KB
/
my_model_selectors.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
import math
import statistics
import warnings
import numpy as np
from hmmlearn.hmm import GaussianHMM
from sklearn.model_selection import KFold
from asl_utils import combine_sequences
class ModelSelector(object):
'''
base class for model selection (strategy design pattern)
'''
def __init__(self, all_word_sequences: dict, all_word_Xlengths: dict, this_word: str,
n_constant=3,
min_n_components=2, max_n_components=10,
random_state=14, verbose=False):
self.words = all_word_sequences
self.hwords = all_word_Xlengths
self.sequences = all_word_sequences[this_word]
self.X, self.lengths = all_word_Xlengths[this_word]
self.this_word = this_word
self.n_constant = n_constant
self.min_n_components = min_n_components
self.max_n_components = max_n_components
self.random_state = random_state
self.verbose = verbose
def select(self):
raise NotImplementedError
def base_model(self, num_states):
# with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=DeprecationWarning)
# warnings.filterwarnings("ignore", category=RuntimeWarning)
try:
hmm_model = GaussianHMM(n_components=num_states, covariance_type="diag", n_iter=1000,
random_state=self.random_state, verbose=False).fit(self.X, self.lengths)
if self.verbose:
print("model created for {} with {} states".format(self.this_word, num_states))
return hmm_model
except:
if self.verbose:
print("failure on {} with {} states".format(self.this_word, num_states))
return None
class SelectorConstant(ModelSelector):
""" select the model with value self.n_constant
"""
def select(self):
""" select based on n_constant value
:return: GaussianHMM object
"""
best_num_components = self.n_constant
return self.base_model(best_num_components)
class SelectorBIC(ModelSelector):
""" select the model with the lowest Baysian Information Criterion(BIC) score
http://www2.imm.dtu.dk/courses/02433/doc/ch6_slides.pdf
Bayesian information criteria: BIC = -2 * logL + p * logN
"""
def select(self):
""" select the best model for self.this_word based on
BIC score for n between self.min_n_components and self.max_n_components
:return: GaussianHMM object
"""
warnings.filterwarnings("ignore", category=DeprecationWarning)
# setting a default model and score at the beginning
bic_score = 99999999999
saved_model = self.base_model(self.n_constant)
# iterating from min to max number of components
for i in range(self.min_n_components,self.max_n_components + 1):
try:
#training the model
hmm_model = GaussianHMM(n_components=i, covariance_type="diag", n_iter=1000,random_state=self.random_state, verbose=False).fit(self.X, self.lengths)
# calculating to find the bic score
logL = hmm_model.score(self.X, self.lengths)
bic1 = -2 * logL
num_data_points = len(self.X)
num_features = len(self.X[0])
num_params = i * i + 2 * i * num_features - 1
bic2 = num_params * math.log(num_data_points)
bic_current = bic1 + bic2
#comparing the bic score with the previous one to look for a better one
if bic_current < bic_score:
saved_model = hmm_model
bic_score = bic_current
except:
#if exception occurs, just continuing to next one
continue
return saved_model
class SelectorDIC(ModelSelector):
''' select best model based on Discriminative Information Criterion
Biem, Alain. "A model selection criterion for classification: Application to hmm topology optimization."
Document Analysis and Recognition, 2003. Proceedings. Seventh International Conference on. IEEE, 2003.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.58.6208&rep=rep1&type=pdf
DIC = log(P(X(i)) - 1/(M-1)SUM(log(P(X(all but i))
'''
def select(self):
warnings.filterwarnings("ignore", category=DeprecationWarning)
# having the default score and model at the beginning
dic_score = -99999999999
saved_model = self.base_model(self.n_constant)
# iterating through min to max number of components to find the best one
for i in range(self.min_n_components,self.max_n_components + 1):
try:
# getting the model
hmm_model = GaussianHMM(n_components=i, covariance_type="diag", n_iter=1000,random_state=self.random_state, verbose=False).fit(self.X, self.lengths)
# getting the log likelihood of the current example
logL_i = hmm_model.score(self.X, self.lengths)
except:
# if fails, will continue to next word
continue
logL_rest = 0
# will get the model of all other examples to calculate dic score
for key in self.words:
if key == self.this_word:
continue
X_temp, lengths_temp = self.hwords[key]
try:
hmm_model_temp = GaussianHMM(n_components=i, covariance_type="diag", n_iter=1000,random_state=self.random_state, verbose=False).fit(X_temp, lengths_temp)
# accumulating log likelihood of all examples
logL_rest += hmm_model_temp.score(X_temp, lengths_temp)
except:
continue
coeff = 1 / (len(self.words) -1 )
dic_current = logL_i - coeff * logL_rest
# comparing for the best score
if dic_current > dic_score:
saved_model = hmm_model
dic_score = dic_current
return saved_model
class SelectorCV(ModelSelector):
''' select best model based on average log Likelihood of cross-validation folds
'''
def select(self):
warnings.filterwarnings("ignore", category=DeprecationWarning)
#Initializing and choosing number of folds
best_logL = -9999999999
split_method = KFold(n_splits=3, shuffle=False, random_state=None)
# Iteraing through min to max components
for i in range(self.min_n_components,self.max_n_components + 1):
logL = 0
try:
# combining the data after choosing the indexes for training and test data
for cv_train_idx, cv_test_idx in split_method.split(self.sequences):
X_train, lengths_train = combine_sequences(cv_train_idx,self.sequences)
X_test, lengths_test = combine_sequences(cv_test_idx,self.sequences)
# getting model and computing log likelihood
hmm_model = GaussianHMM(n_components=i, covariance_type="diag", n_iter=1000,random_state=self.random_state, verbose=False).fit(X_train, lengths_train)
logL += hmm_model.score(X_test, lengths_test)
logL = logL/3
# comparing for best model
if logL > best_logL :
best_logL = logL
best_model = GaussianHMM(n_components=i, covariance_type="diag", n_iter=1000,random_state=self.random_state, verbose=False).fit(self.X,self.lengths)
except:
# return default model in case of exception
return self.base_model(self.n_constant)
return best_model