-
Notifications
You must be signed in to change notification settings - Fork 0
/
3.2.dl_tf_advance_minimise_overfitting.py
248 lines (191 loc) · 10.8 KB
/
3.2.dl_tf_advance_minimise_overfitting.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
#%% First standard without any optimization Technique
import numpy as np
import tensorflow as tf
import os
from pandas_ml import ConfusionMatrix
import matplotlib.pyplot as plt
import gc; gc.enable()
os.chdir("D:\\trainings\\tensorflow")
tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.INFO)
# fix random seed for reproducibility
seed = 123; np.random.seed(seed); tf.compat.v1.set_random_seed(seed)
#List of avalble TF dataset https://www.tensorflow.org/versions/r2.0/api_docs/python/tf/keras/datasets?hl=en
#Details is at http://yann.lecun.com/exdb/mnist/
#https://github.com/petar/GoMNIST/tree/master/data
# Load training and eval data
(train_data, train_labels), (eval_data, eval_labels) = tf.keras.datasets.mnist.load_data()
train_data.shape # (60000, 28, 28)
eval_data.shape # (10000, 28, 28)
train_labels
eval_labels
#constants
im_wh = train_data.shape[1] # Assuming that width and height are same else make by transformation
#View the first 10 images and the class name
plt.figure(figsize=(10,10))
for i in range(10):
plt.subplot(2,5,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
x_image = np.reshape(train_data[i], [im_wh, im_wh])
plt.imshow(x_image, cmap=plt.cm.binary)
plt.xlabel(train_labels[i])
# end of 'for'
#Reshape data to add channel as per need of CNN - (image_height, image_width, color_channels)
train_data = train_data.reshape((train_data.shape[0], im_wh, im_wh, 1))
eval_data = eval_data.reshape((eval_data.shape[0], im_wh, im_wh, 1))
#Scale these values to a range of 0 to 1
train_data = train_data / 255.0
eval_data = eval_data / 255.0
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', input_shape=(im_wh, im_wh, 1))) #, padding="same"
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu'))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# to keep time short, set epochs = 1
model.fit(train_data, train_labels, epochs=1, workers=4, use_multiprocessing=True)
#Making Predictions
predictions = model.predict(eval_data)
# Extracting max probability
predictions_number = np.array([])
for row_num in range(predictions.shape[0]): # row_num = 0
predictions_number = np.append(predictions_number, np.argmax(predictions[row_num]))
# Just precaution
predictions_number = predictions_number.astype(int)
confusion_matrix = ConfusionMatrix(eval_labels, predictions_number)
cms = confusion_matrix.stats()
print("Overall Accuracy is ", round(cms['overall']['Accuracy'], 2),", Kappa is ", round(cms['overall']['Kappa'], 2))
# none: Overall Accuracy is 0.99 , Kappa is 0.99
#%% Weight regularization: L2 (weight decay) -> cost added is proportional to the square of the value of the weights coefficients
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', input_shape=(im_wh, im_wh, 1), kernel_regularizer= tf.keras.regularizers.l2(0.001)))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', kernel_regularizer= tf.keras.regularizers.l2(0.001)))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu', kernel_regularizer= tf.keras.regularizers.l2(0.001)))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# to keep time short, set epochs = 1
model.fit(train_data, train_labels, epochs=1, workers=4, use_multiprocessing=True)
#Making Predictions
predictions = model.predict(eval_data)
# Extracting max probability
predictions_number = np.array([])
for row_num in range(predictions.shape[0]): # row_num = 0
predictions_number = np.append(predictions_number, np.argmax(predictions[row_num]))
# Just precaution
predictions_number = predictions_number.astype(int)
confusion_matrix = ConfusionMatrix(eval_labels, predictions_number)
cms = confusion_matrix.stats()
print("Overall Accuracy is ", round(cms['overall']['Accuracy'], 2),", Kappa is ", round(cms['overall']['Kappa'], 2))
# L2: Overall Accuracy is 0.98 , Kappa is 0.98
# none: Overall Accuracy is 0.99 , Kappa is 0.99
#%% Weight regularization: L1 -> cost added is proportional to the absolute value of the weights coefficients
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', input_shape=(im_wh, im_wh, 1), kernel_regularizer= tf.keras.regularizers.l1(0.001)))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', kernel_regularizer= tf.keras.regularizers.l1(0.001)))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu', kernel_regularizer= tf.keras.regularizers.l1(0.001)))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# to keep time short, set epochs = 1
model.fit(train_data, train_labels, epochs=1, workers=4, use_multiprocessing=True)
#Making Predictions
predictions = model.predict(eval_data)
# Extracting max probability
predictions_number = np.array([])
for row_num in range(predictions.shape[0]): # row_num = 0
predictions_number = np.append(predictions_number, np.argmax(predictions[row_num]))
# Just precaution
predictions_number = predictions_number.astype(int)
confusion_matrix = ConfusionMatrix(eval_labels, predictions_number)
cms = confusion_matrix.stats()
print("Overall Accuracy is ", round(cms['overall']['Accuracy'], 2),", Kappa is ", round(cms['overall']['Kappa'], 2))
# L1: Overall Accuracy is 0.97 , Kappa is 0.96
# L2: Overall Accuracy is 0.98 , Kappa is 0.98
# none: Overall Accuracy is 0.99 , Kappa is 0.99
#%% Dropout
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', input_shape=(im_wh, im_wh, 1)))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Dropout(0.25)) # https://stats.stackexchange.com/questions/147850/are-pooling-layers-added-before-or-after-dropout-layers
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu'))
model.add(tf.keras.layers.Dropout(0.25))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dropout(0.25))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# to keep time short, set epochs = 1
model.fit(train_data, train_labels, epochs=1, workers=4, use_multiprocessing=True)
#Making Predictions
predictions = model.predict(eval_data)
# Extracting max probability
predictions_number = np.array([])
for row_num in range(predictions.shape[0]): # row_num = 0
predictions_number = np.append(predictions_number, np.argmax(predictions[row_num]))
# Just precaution
predictions_number = predictions_number.astype(int)
confusion_matrix = ConfusionMatrix(eval_labels, predictions_number)
cms = confusion_matrix.stats()
print("Overall Accuracy is ", round(cms['overall']['Accuracy'], 2),", Kappa is ", round(cms['overall']['Kappa'], 2))
# Dropout: Overall Accuracy is 0.99 , Kappa is 0.99
# L1: Overall Accuracy is 0.97 , Kappa is 0.96
# L2: Overall Accuracy is 0.98 , Kappa is 0.98
# none: Overall Accuracy is 0.99 , Kappa is 0.99
#%% Batch Normalisation
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), input_shape=(im_wh, im_wh, 1))) # , activation='relu'
model.add(tf.keras.layers.BatchNormalization())
model.add(tf.keras.layers.ReLU()) # It is advisable to add activation after BN
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (5, 5)))
model.add(tf.keras.layers.BatchNormalization())
model.add(tf.keras.layers.ReLU()) # It is advisable to add activation after BN
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64))
model.add(tf.keras.layers.BatchNormalization())
model.add(tf.keras.layers.ReLU()) # It is advisable to add activation after BN
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# to keep time short, set epochs = 1
model.fit(train_data, train_labels, epochs=1, workers=4, use_multiprocessing=True)
#Making Predictions
predictions = model.predict(eval_data)
# Extracting max probability
predictions_number = np.array([])
for row_num in range(predictions.shape[0]): # row_num = 0
predictions_number = np.append(predictions_number, np.argmax(predictions[row_num]))
# Just precaution
predictions_number = predictions_number.astype(int)
confusion_matrix = ConfusionMatrix(eval_labels, predictions_number)
cms = confusion_matrix.stats()
print("Overall Accuracy is ", round(cms['overall']['Accuracy'], 2),", Kappa is ", round(cms['overall']['Kappa'], 2))
# Batch Normalisation: Overall Accuracy is 0.99 , Kappa is 0.99
# Dropout: Overall Accuracy is 0.99 , Kappa is 0.99
# L1: Overall Accuracy is 0.97 , Kappa is 0.96
# L2: Overall Accuracy is 0.98 , Kappa is 0.98
# none: Overall Accuracy is 0.99 , Kappa is 0.99
#%% EarlyStopping
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu', input_shape=(im_wh, im_wh, 1))) #, padding="same"
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (5, 5), activation='relu'))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.summary() # Note shape of images is going down. Note: it'll not go down when padding="same"
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
#Condition, when to early stop
early_stop = tf.keras.callbacks.EarlyStopping(monitor='accuracy', patience=1)
# to keep time short, set epochs = 1 & [:5000]
model.fit(train_data[:5000], train_labels[:5000], epochs=100, workers=4, use_multiprocessing=True, callbacks=[early_stop])