1 import tensorflow as tf
2 from tensorflow.keras.layers import Dense
3 from tensorflow.keras import Model
4 from sklearn import datasets
5 import numpy as np
6
7
8 x_train = datasets.load_iris().data
9 y_train = datasets.load_iris().target
10
11
12 np.random.seed(116)
13 np.random.shuffle(x_train)
14 np.random.seed(116)
15 np.random.shuffle(y_train)
16 tf.random.set_seed(116)
17
18
19
20 class IrisModel(Model):
21 def __init__(self):
22 super(IrisModel, self).__init__()
23 self.d1 = Dense(3, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2())
24
25 def call(self, x):
26 y = self.d1(x)
27 return y
28
29 model = IrisModel()
30
31
32 model.compile(optimizer=tf.keras.optimizers.SGD(lr=0.1),
33 loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
34 metrics=['sparse_categorical_accuracy'])
35
36 model.fit(x_train, y_train, batch_size=32, epochs=500, validation_split=0.2, validation_freq=20)
37
38 model.summary()
