In this tutorial,we will use some examples to show you how to use tf.layers.max_pooling1d() function in tensorflow.
Syntax
tf.layers.max_pooling1d() is defined as:
tf.layers.max_pooling1d(
inputs,
pool_size,
strides,
padding='valid',
data_format='channels_last',
name=None
)It is used for max pooling layer for 1D inputs on length axis.
Parameters
inputs: tensor with rank 3, [batch_size, length, channels]
pool_size: the size of the pooling window, it can be integer or tuple, list
strides: the strides of the pooling operation, integer or tuple/list
padding: valid or same
It will also return a tensor with rank 3.
How to use tf.layers.max_pooling1d()?
We will use an example to show you how to do.
For example:
import tensorflow as tf
a=tf.constant([[[0.1,0.2,0.5,0.2],
[0.11,0.1,0.1,0.7]],
[[0.2,0.33,0.21,0.43],
[0.1,0.12,0.22,0.4]],
[[0.2,0.3,0.2,0.3],
[0.12,0.2,0.2,0.4]]])
print(a)
v = tf.layers.max_pooling1d(a, pool_size=2, strides=1, padding='same')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
x = sess.run(v)
print(x.shape)
print(x)Here a is [3, 2, 4], when pool_size=2, strides=1, padding=’same’, we will get:
(3, 2, 4) [[[0.11 0.2 0.5 0.7 ] [0.11 0.2 0.5 0.7 ]] [[0.2 0.33 0.22 0.43] [0.2 0.33 0.22 0.43]] [[0.2 0.3 0.2 0.4 ] [0.2 0.3 0.2 0.4 ]]]
From the above, we can find length = 2, it is equal to pool_size, length = pool_size = 2.
How about pool_size > length?
v = tf.layers.max_pooling1d(a, pool_size=3, strides=1, padding='same')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
x = sess.run(v)
print(x.shape)
print(x)
Here pool_size = 3 > length = 2, we will see:
(3, 2, 4) [[[0.11 0.2 0.5 0.7 ] [0.11 0.2 0.5 0.7 ]] [[0.2 0.33 0.22 0.43] [0.2 0.33 0.22 0.43]] [[0.2 0.3 0.2 0.4 ] [0.2 0.3 0.2 0.4 ]]]
The result is same to pool_size = 2.
How about pool_size > length when padding = ‘valid’?
Look at this example:
v = tf.layers.max_pooling1d(a, pool_size=3, strides=1, padding='valid')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
x = sess.run(v)
print(x.shape)
print(x)
Run this code, we will get: ValueError: Negative dimension size caused by subtracting 3 from 2 for ‘max_pooling1d/MaxPool’ (op: ‘MaxPool’) with input shapes: [3,2,1,4].
tf.layers.max_pooling1d() vs tf.nn.max_pool()?
tf.nn.max_pool() also can implement max pool in tensorflow. Here is the tutorial:
tf.nn.max_pool() also can work as tf.layers.max_pooling1d(), for example:
import tensorflow as tf
import numpy as np
a=tf.constant([[[0.1,0.2,0.5,0.2],
[0.11,0.1,0.1,0.7],
[0.01,0.12,0.25,0.12]],
[[0.2,0.33,0.21,0.43],
[0.1,0.23,0.11,0.53],
[0.1,0.12,0.22,0.4]],
[[0.2,0.3,0.2,0.3],
[0.22,0.13,0.12,0.13],
[0.12,0.2,0.2,0.4]]])
print(a)
v = tf.layers.max_pooling1d(a, pool_size=2, strides=1, padding='SAME')
a1 = tf.reshape(a,[3, 3, 1, 4])
v2 = tf.nn.max_pool(a1, ksize = [1, 2, 1, 1], strides = [1, 1, 1, 1], padding = 'SAME', data_format='NHWC')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
x = sess.run(v)
print(x.shape)
print(x)
x2 = sess.run(v2)
x2 = np.reshape(x2, [3, 3, 4])
print(x2.shape)
print(x2)
Run this code, we will get:
(3, 3, 4) [[[0.11 0.2 0.5 0.7 ] [0.11 0.12 0.25 0.7 ] [0.01 0.12 0.25 0.12]] [[0.2 0.33 0.21 0.53] [0.1 0.23 0.22 0.53] [0.1 0.12 0.22 0.4 ]] [[0.22 0.3 0.2 0.3 ] [0.22 0.2 0.2 0.4 ] [0.12 0.2 0.2 0.4 ]]] (3, 3, 4) [[[0.11 0.2 0.5 0.7 ] [0.11 0.12 0.25 0.7 ] [0.01 0.12 0.25 0.12]] [[0.2 0.33 0.21 0.53] [0.1 0.23 0.22 0.53] [0.1 0.12 0.22 0.4 ]] [[0.22 0.3 0.2 0.3 ] [0.22 0.2 0.2 0.4 ] [0.12 0.2 0.2 0.4 ]]]
x and x2 are the same.