关于activation(bn)

Question

activation(bn)
为什么可以直接可以这样用，activation这个函数是怎么来的，它不属于tf吗

Answer 1

正十七 2020-03-05 22:05:49

可以看代码：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

def conv_wrapper(inputs,                   name,                   is_training,                  output_channel = 32,                  kernel_size = (3,3),                  activation = tf.nn.relu,                  padding = 'same'):     """wrapper of tf.layers.conv2d"""     # without bn: conv -> activation     # with batch normalization: conv -> bn -> activation     with tf.name_scope(name):         conv2d = tf.layers.conv2d(inputs,                                   output_channel,                                   kernel_size,                                   padding = padding,                                   activation = None,                                   name = name + '/conv2d')         bn = tf.layers.batch_normalization(conv2d,                                            training = is_training)         return activation(bn)

这里的activation是tensorflow里的激活函数啊，只不过它是从外面传进来的。默认是tf.nn.relu.