第三讲 神经网络八股--自定义class model 分类iris 

 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()

 

posted @ 2020-05-04 21:38  WWBlog  阅读(362)  评论(0)    收藏  举报