-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnn_models_CIFAR100.py
239 lines (145 loc) · 7.24 KB
/
nn_models_CIFAR100.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
import random
import math
import matplotlib.pyplot as plt
from matplotlib.pyplot import figure
import scipy
from scipy.special import softmax
import numpy as np
# Typing
import typing
from typing import TypeVar, Generic
from collections.abc import Callable
from tqdm import tqdm
from collections import namedtuple
import statistics
import dataclasses
from dataclasses import dataclass
from tensorflow.keras.optimizers import Adam
from tensorflow.keras import datasets, layers, models
#import keras.backend as K
import copy
from copy import deepcopy
import tensorflow as tf
NN_Individual = namedtuple("NN_Individual", ["nn", "opt_obj", "LR_constant", "reg_constant"])
## CIFAR 100
# Testing population descent
def new_CIFAR100_individual_without_regularization():
# model #6, no_reg - better, bigger CIFAR10 model
model_num = "1 no_reg CIFAR100"
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, kernel_size = 3, activation='relu', input_shape = (32, 32, 3)),
# tf.keras.layers.BatchNormalization(),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Conv2D(64, kernel_size = 3, strides=1, activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(128, kernel_size = 3, strides=1, padding='same', activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(64, kernel_size = 3, activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((4, 4)),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(256, activation = "relu"),
tf.keras.layers.Dense(100, activation = "softmax")
])
optimizer = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3) # 1e-3 (for FMNIST)
LR_constant = 10**(np.random.normal(-4, 2))
reg_constant = 10**(np.random.normal(0, 2))
# creating NN object with initialized parameters
NN_object = NN_Individual(model, optimizer, LR_constant, reg_constant)
return NN_object, model_num
# Testing population descent
def new_CIFAR100_individual_with_regularization():
# model #6, no_reg - better, bigger CIFAR10 model
model_num = "1 w/ reg CIFAR100"
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, kernel_size = 3, activation='relu', input_shape = (32, 32, 3), kernel_regularizer=tf.keras.regularizers.l2(l=.001)),
# tf.keras.layers.BatchNormalization(),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Conv2D(64, kernel_size = 3, strides=1, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=.001)),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(128, kernel_size = 3, strides=1, padding='same', activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=.001)),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(64, kernel_size = 3, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=.001)),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((4, 4)),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(256, activation = "relu", kernel_regularizer=tf.keras.regularizers.l2(l=.001)),
tf.keras.layers.Dense(100, activation = "softmax")
])
optimizer = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3) # 1e-3 (for FMNIST)
LR_constant = 10**(np.random.normal(-4, 2))
reg_constant = 10**(np.random.normal(0, 2))
# creating NN object with initialized parameters
NN_object = NN_Individual(model, optimizer, LR_constant, reg_constant)
return NN_object, model_num
# Testing Hyperparameter search
def new_CIFAR_hps_individual_without_regularization():
regularization_amount = [0]
learning_rate = [0.01, 0.001, 0.0001, 0.00001, 0.000001]
population = []
reg_list = []
for r in range(len(regularization_amount)):
for l in range(len(learning_rate)):
model_num = "6 CIFAR without_reg"
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, kernel_size = 3, activation='relu', input_shape = (32, 32, 3)),
# tf.keras.layers.BatchNormalization(),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Conv2D(64, kernel_size = 3, strides=1, activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(128, kernel_size = 3, strides=1, padding='same', activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(64, kernel_size = 3, activation='relu'),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((4, 4)),
# tf.keras.layers.Dropout(0.2),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(256, activation = "relu", kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
tf.keras.layers.Dense(100, activation = "softmax")
])
optimizer = tf.keras.optimizers.legacy.Adam(learning_rate=learning_rate[l])
model.compile(optimizer=optimizer,
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
population.append(model)
reg_list.append(regularization_amount[r])
population = np.array(population)
return population, reg_list, model_num
# Testing Hyperparameter search
def new_CIFAR_hps_individual_with_regularization():
learning_rate = [0.01, 0.001, 0.0001, 0.00001, 0.000001]
regularization_amount = [0.01, 0.001, 0.0001, 0.00001, 0.000001]
population = []
reg_list = []
for r in range(len(regularization_amount)):
for l in range(len(learning_rate)):
model_num = "6 CIFAR with_reg"
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, kernel_size = 3, activation='relu', input_shape = (32, 32, 3), kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
tf.keras.layers.Conv2D(64, kernel_size = 3, strides=1, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(128, kernel_size = 3, strides=1, padding='same', activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(64, kernel_size = 3, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
tf.keras.layers.MaxPooling2D((4, 4)),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(256, activation = "relu", kernel_regularizer=tf.keras.regularizers.l2(l=regularization_amount[r])),
tf.keras.layers.Dense(100, activation = "softmax")
])
optimizer = tf.keras.optimizers.legacy.Adam(learning_rate=learning_rate[l])
model.compile(optimizer=optimizer,
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
population.append(model)
reg_list.append(regularization_amount[r])
population = np.array(population)
return population, reg_list, model_num