-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
183 lines (141 loc) · 6.38 KB
/
main.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
# -*- coding: utf-8 -*-
# Created by Thais Luca
# Systems Engineering and Computer Science Program - COPPE, Federal University of Rio de Janeiro
# Created in 05/01/2020
# Last update in 06/05/2020
import wisardpkg as wp
import pandas as pd
import numpy as np
import sys, os
import matplotlib.pyplot as plt
import seaborn as sn
from collections import Counter
from pandas import compat
compat.PY3 = True
pd.options.display.float_format = '{:.2f}'.format
from sklearn.model_selection import train_test_split, KFold
from sklearn.metrics import confusion_matrix, accuracy_score
def load_dataset():
dataset = pd.read_csv('dataset/mnist_train.csv')
dataset = pd.concat([dataset, pd.read_csv('dataset/mnist_test.csv')])
return dataset
def preprocess(df, threshold):
columns = df.columns
for column in columns:
df[column] = np.where(df[column] >= threshold, 1, 0)
return df
def cm2df(cm, labels):
df = pd.DataFrame()
# rows
for i, row_label in enumerate(labels):
rowdata={}
# columns
for j, col_label in enumerate(labels):
rowdata[col_label]=cm[i,j]
df = df.append(pd.DataFrame.from_dict({row_label:rowdata}, orient='index'))
return df
def save_accuracy(threshold, addressSize, train_accuracy_score, validation_accuracy_score, test_accuracy_score):
train_accuracy_score_mean = np.mean(train_accuracy_score, axis=0)
validation_accuracy_score_mean = np.mean(validation_accuracy_score, axis=0)
test_accuracy_score_mean = np.mean(test_accuracy_score, axis=0)
train_accuracy_score_std = np.std(train_accuracy_score)
validation_accuracy_score_std = np.std(validation_accuracy_score)
test_accuracy_score_std = np.std(test_accuracy_score)
matrix = {}
matrix['threshold'] = [threshold]
matrix['addressSize'] = [addressSize]
matrix['train_accuracy_mean'] = [train_accuracy_score_mean]
matrix['validation_accuracy_mean'] = [validation_accuracy_score_mean]
matrix['test_accuracy_mean'] = [test_accuracy_score_mean]
matrix['train_accuracy_std'] = [train_accuracy_score_std]
matrix['validation_accuracy_std'] = [validation_accuracy_score_std]
matrix['test_accuracy_std'] = [test_accuracy_score_std]
result = pd.DataFrame(matrix)
with open('results/accuracy.csv', 'a') as file:
result.to_csv(file, index=False, header=True)
def save_matrix(cm, filename):
df = cm2df(cm, range(10))
with open(filename, 'a') as file:
df.to_csv(file)
def plot_heatmap(cm, title, filename):
df_cm = pd.DataFrame(cm, index = range(10), columns = range(10))
plt.figure(figsize = (10,7))
sn.heatmap(df_cm, cmap="YlGnBu", annot=True)
plt.title(title)
plt.ylabel('Target Value')
plt.xlabel('Predicted Value')
plt.savefig('results/' + filename + '.jpg')
#plt.show()
def main(input_threshold, input_addressSize):
confusion_matrix_train_scores = np.zeros((10,10))
confusion_matrix_validation_scores = np.zeros((10,10))
confusion_matrix_test_scores = np.zeros((10,10))
train_accuracy_score = []
validation_accuracy_score = []
test_accuracy_score = []
addressSize = input_addressSize # number of addressing bits in the ram
ignoreZero = False # optional; causes the rams to ignore the address 0
n_splits = 10 # number of splits used in KFold
threshold = input_threshold
# False by default for performance reasons,
# when True, WiSARD prints the progress of train() and classify()
verbose = True
dataset = load_dataset()
train, test = train_test_split(dataset, test_size=0.3)
print(len(train), len(test))
X = preprocess(train.drop(['label'], axis=1), threshold).values.tolist()
Y = train['label'].values.tolist()
X_test = preprocess(test.drop(['label'], axis=1), threshold).values.tolist()
Y_test = test['label'].values.tolist()
Y_test = [str(y) for y in Y_test]
# Define model
wsd = wp.Wisard(addressSize, ignoreZero=ignoreZero, verbose=verbose)
kf = KFold(n_splits=n_splits, shuffle=True)
fold = 1
for train_index, val_index in kf.split(X):
print("FOLD:", fold)
print("TRAIN: {} VALIDATION: {}".format(len(train_index), len(val_index)))
x_train, x_val = [X[index] for index in train_index], [X[index] for index in val_index]
y_train, y_val = [str(Y[index]) for index in train_index], [str(Y[index]) for index in val_index]
#print("Training: ", Counter(y_train))
#print("Validation: ", Counter(y_val))
#print("Test: ", Counter(Y_test))
# train using the input data
print("Training")
wsd.train(x_train,y_train)
# classify train data
print("Train data classification")
out_train = wsd.classify(x_train)
cm_training = confusion_matrix(y_train, out_train)
cm_training = cm_training / cm_training.astype(np.float).sum(axis=1)
confusion_matrix_train_scores += cm_training
train_accuracy_score.append(accuracy_score(y_train, out_train))
# classify validation data
print("Validation data classification")
out_val = wsd.classify(x_val)
cm_validation = confusion_matrix(y_val, out_val)
cm_validation = cm_validation / cm_validation.astype(np.float).sum(axis=1)
confusion_matrix_validation_scores += cm_validation
validation_accuracy_score.append(accuracy_score(y_val, out_val))
#classify test data
print("Test data classification")
out_test = wsd.classify(X_test)
cm_test = confusion_matrix(Y_test, out_test)
cm_test = cm_test / cm_test.astype(np.float).sum(axis=1)
confusion_matrix_test_scores += cm_test
test_accuracy_score.append(accuracy_score(Y_test, out_test))
fold += 1
print('\n')
confusion_matrix_train_scores = np.divide(confusion_matrix_train_scores, n_splits)
confusion_matrix_validation_scores = np.divide(confusion_matrix_validation_scores, n_splits)
confusion_matrix_test_scores = np.divide(confusion_matrix_test_scores, n_splits)
save_accuracy(threshold, addressSize, train_accuracy_score, validation_accuracy_score, test_accuracy_score)
path = 'results/threshold_' + str(threshold) + 'addressSize_' + str(addressSize)
if(not os.path.isdir(path)):
os.mkdir(path)
plot_heatmap(confusion_matrix_train_scores, title='Training', filename=path + '/training_heatmap')
plot_heatmap(confusion_matrix_validation_scores, title='Validation', filename=path + '/validation_heatmap')
plot_heatmap(confusion_matrix_test_scores, title='Test', filename=path + '/test_heatmap')
save_matrix(confusion_matrix_train_scores, path + '/training_confusion_matrix.csv')
save_matrix(confusion_matrix_validation_scores, path + '/validation_confusion_matrix.csv')
save_matrix(confusion_matrix_test_scores, path + '/test_confusion_matrix.csv')