-
Notifications
You must be signed in to change notification settings - Fork 0
/
invariantMnistExample.py
38 lines (30 loc) · 1.16 KB
/
invariantMnistExample.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
#Example using polynomial layer in a residual network. Not particularly great
#on this problem, but it may do better on time series.
import tensorflow as tf
from high_order_layers import PolynomialLayers as poly
from tensorflow.keras.layers import *
mnist = tf.keras.datasets.mnist
layers = tf.keras.layers
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = (x_train / 128.0 - 1.0), (x_test / 128.0 - 1.0)
units = 10
basis = poly.b3
#Residual layer
def res_block(input_data, units=units, basis=basis) :
x0 = LayerNormalization()(input_data)
x1 = poly.Polynomial(units, basis=basis)(x0)
x1 = Add()([x1, input_data])
return x1
inputs = tf.keras.Input(shape=(28,28))
x = Flatten(input_shape=(28, 28))(inputs)
x = poly.Polynomial(units, basis=basis)(x)
for i in range(3) :
x = res_block(x, basis=basis, units=units)
x = LayerNormalization()(x)
outputs = Dense(10, activation='softmax')(x)
model = tf.keras.Model(inputs, outputs)
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
model.fit(x_train, y_train, epochs=5, batch_size=10)
model.evaluate(x_test, y_test)