1 tf.nn.max_pool(
2     value,
3     ksize,
4     strides,
5     padding,
6     data_format='NHWC',
7     name=None
8 )

Pooling is similar to convolution, The reason is that pooling is similar to subsampling

1 tf.layers.max_pooling2d(
2     inputs,
3     pool_size,
4     strides,
5     padding='valid',
6     data_format='channels_last',
7     name=None
8 )