-
Notifications
You must be signed in to change notification settings - Fork 0
/
apply_classifier.py
185 lines (131 loc) · 7.42 KB
/
apply_classifier.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
from matplotlib.pyplot import cla
import numpy as np
import pandas as pd
import sys
import os
import tensorflow as tf
from classifiers.KNN import KNN_Classifier
from classifiers.SVM import SVM_Classifier
from classifiers.MLP import MLP_Classifier
from classifiers.RIDGE import Ridge_Classifier
from utils.utils import load_data, znormalisation, encode_labels
from utils.concat_supervised import concatenate_supervised_unsupervised
from sklearn.metrics import accuracy_score
gpus = tf.config.list_physical_devices('GPU')
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
# returns score
def extract_args():
if len(sys.argv) != 7:
raise ValueError("No options were specified")
else:
return sys.argv[2], sys.argv[4], sys.argv[6]
# example : python3 main.py -e inception -o results/ -d Coffee
def apply_classifier(classifier_name,xtrain,ytrain,xtest,ytest):
if classifier_name == 'knn':
clf = KNN_Classifier()
elif classifier_name == 'mlp':
clf = MLP_Classifier(xtrain=xtrain,ytrain=ytrain,lr_0=0.1,epochs=2000)
elif classifier_name == 'svm':
clf = SVM_Classifier()
elif classifier_name == 'ridge':
clf = Ridge_Classifier()
else:
raise ValueError("no such classifier as "+classifier_name)
if classifier_name != 'mlp':
clf.fit(xtrain=xtrain,ytrain=ytrain)
elif classifier_name == 'mlp':
clf.fit()
ypred = clf.predict(xtest=xtest)
return accuracy_score(y_true=ytest,y_pred=ypred,normalize=True)
if __name__ == "__main__":
encoder_name, output_directory, file_name = extract_args()
print(file_name)
xtrain, ytrain, xtest, ytest = load_data(file_name=file_name)
ytrain = encode_labels(ytrain)
ytest = encode_labels(ytest)
runs = 5
exps = 5
if output_directory == 'results/':
if os.path.exists(output_directory+encoder_name+'/results_ucr.csv'):
df = pd.read_csv(output_directory+encoder_name+'/results_ucr.csv')
else:
df = pd.DataFrame(columns=['encoder','dataset','1-NN','1-NN-std','SVM','SVM-std',
'RIDGE','RIDGE-std','1-LP','1-LP-std',
'concatenate-supervised-1-LP','concatenate-supervised-1-LP-std'])
Score_knn = []
Score_svm = []
Score_ridge = []
Score_mlp = []
Score_concat = []
for _run in range(runs):
vtrain = np.load(output_directory+encoder_name+'/run_'+str(_run)+'/'+file_name+'/v_train.npy')
vtest = np.load(output_directory+encoder_name+'/run_'+str(_run)+'/'+file_name+'/v_test.npy')
Score_knn.append(apply_classifier(classifier_name='knn',xtrain=vtrain,ytrain=ytrain,
xtest=vtest,ytest=ytest))
Score_svm.append(apply_classifier(classifier_name='svm',xtrain=vtrain,ytrain=ytrain,
xtest=vtest,ytest=ytest))
Score_ridge.append(apply_classifier(classifier_name='ridge',xtrain=vtrain,ytrain=ytrain,
xtest=vtest,ytest=ytest))
Score_mlp.append(apply_classifier(classifier_name='mlp',xtrain=vtrain,ytrain=ytrain,
xtest=vtest,ytest=ytest))
tf.keras.backend.clear_session()
path_model_supervised = 'supervised/'+encoder_name+'/run_'+str(_run)+'/'+file_name+'/best_model.hdf5'
new_xtrain, new_xtest = concatenate_supervised_unsupervised(path_model_supervised=path_model_supervised,
xtrain=xtrain,xtest=xtest,
vtrain=vtrain,vtest=vtest)
Score_concat.append(apply_classifier(classifier_name='mlp',xtrain=new_xtrain,ytrain=ytrain,xtest=new_xtest,ytest=ytest))
tf.keras.backend.clear_session()
df = df.append({
'encoder' : encoder_name,
'dataset' : file_name,
'1-NN' : np.mean(Score_knn),
'1-NN-std' : np.std(Score_knn),
'SVM' : np.mean(Score_svm),
'SVM-std' : np.std(Score_svm),
'RIDGE' : np.mean(Score_ridge),
'RIDGE-std' : np.std(Score_ridge),
'1-LP' : np.mean(Score_mlp),
'1-LP-std' : np.std(Score_mlp),
'concatenate-supervised-1-LP' : np.mean(Score_concat),
'concatenate-supervised-1-LP-std' : np.std(Score_concat)
},ignore_index=True)
df.to_csv(output_directory+encoder_name+'/results_ucr.csv',index=False)
elif output_directory[:-4] == 'results_semi':
if os.path.exists(output_directory+encoder_name+'/results_ucr.csv'):
df = pd.read_csv(output_directory+encoder_name+'/results_ucr.csv')
else:
df = pd.DataFrame(columns=['encoder','dataset','RIDGE-semi','RIDGE-std-semi','RIDGE','RIDGE-std'])
Score_knn = []
Score_svm = []
Score_ridge = []
Score_ridge_semi = []
Score_mlp = []
for _run in range(runs):
v_train = np.load('results/'+encoder_name+'/run_'+str(_run)+'/'+file_name+'/v_train.npy')
vtest = np.load('results/'+encoder_name+'/run_'+str(_run)+'/'+file_name+'/v_test.npy')
for _exp in range(exps):
vtrain_semi = np.load(output_directory+encoder_name+'/run_'+str(_run)+'/'+file_name+'/exp_'+str(_exp)+'/v_train.npy')
vtest_semi = np.load(output_directory+encoder_name+'/run_'+str(_run)+'/'+file_name+'/exp_'+str(_exp)+'/v_test.npy')
semi_indices = np.load(output_directory+encoder_name+'/run_'+str(_run)+'/'+file_name+'/exp_'+str(_exp)+'/train_indices.npy')
semi_ytrain = ytrain[semi_indices]
vtrain = v_train[semi_indices]
# Score_knn.append(apply_classifier(classifier_name='knn',xtrain=vtrain,ytrain=semi_ytrain,
# xtest=vtest,ytest=ytest))
# Score_svm.append(apply_classifier(classifier_name='svm',xtrain=vtrain,ytrain=semi_ytrain,
# xtest=vtest,ytest=ytest))
Score_ridge_semi.append(apply_classifier(classifier_name='ridge',xtrain=vtrain_semi,ytrain=semi_ytrain,
xtest=vtest_semi,ytest=ytest))
Score_ridge.append(apply_classifier(classifier_name='ridge',xtrain=vtrain,ytrain=semi_ytrain,
xtest=vtest,ytest=ytest))
# Score_mlp.append(apply_classifier(classifier_name='mlp',xtrain=vtrain,ytrain=semi_ytrain,
# xtest=vtest,ytest=ytest))
df = df.append({
'encoder' : encoder_name,
'dataset' : file_name,
'RIDGE-semi' : np.mean(Score_ridge_semi),
'RIDGE-std-semi' : np.std(Score_ridge_semi),
'RIDGE' : np.mean(Score_ridge),
'RIDGE-std' : np.std(Score_ridge),
},ignore_index=True)
df.to_csv(output_directory+encoder_name+'/results_ucr.csv',index=False)