PoolingLayer

PoolingLayer[sz]

represents a pooling net layer using kernels of size sz.

PoolingLayer[{w}]

represents a layer performing one-dimensional pooling with kernels of size w.

PoolingLayer[{h,w}]

represents a layer performing two-dimensional pooling with kernels of size h×w.

PoolingLayer[{h,w,d}]

represents a layer performing three-dimensional pooling with kernels of size h×w×d.

PoolingLayer[kernel,stride]

represents a layer that uses stride as the step size between kernel applications.

PoolingLayer[kernel,opts]

includes options for other pooling methods, padding and other parameters.

Details and Options

  • PoolingLayer[n,…] represents a layer that, applied to an input array with c input channels and one or more spatial dimensions, performs c distinct pooling operations across the spatial dimensions to produce an output array with c channels.
  • PoolingLayer[n] is equivalent to PoolingLayer[n,1].
  • The following optional parameters can be included:
  • "Dimensionality"Automaticnumber of spatial dimensions of the pooling
    "Function" Maxaggregation function to use
    Interleaving Falsethe position of the channel dimension
    PaddingSize 0amount of zero padding to apply to the input
  • With the setting InterleavingFalse, the channel dimension is taken to be the first dimension of the input and output arrays.
  • With the setting InterleavingTrue, the channel dimension is taken to be the last dimension of the input and output arrays.
  • The settings for kernel and stride can be of the following forms:
  • nuse the value n for all dimensions
    {…,ni,…}use the value ni for the i^(th) dimension
  • The setting for PaddingSize can be of the following forms:
  • npad every dimension with n zeros on the beginning and end
    {n1,n2,…}pad the i^(th) dimension with n zeros on the beginning and end
    {{n1,m1},{n2,m2},…}pad the i^(th) dimension with ni zeros at the beginning and mi zeros at the end
    "Same"
  • pad every dimension so that the output size is equal to the input size divided by the stride (rounded up)
  • PoolingLayer[…][input] explicitly computes the output from applying the layer.
  • PoolingLayer[…][{input1,input2,…}] explicitly computes outputs for each of the inputi.
  • When given a NumericArray as input, the output will be a NumericArray.
  • PoolingLayer is typically used inside NetChain, NetGraph, etc.
  • NetExtract can be used to extract parameter values from a PoolingLayer object.
  • PoolingLayer exposes the following ports for use in NetGraph etc.:
  • "Input"an array of rank 2, 3 or 4
    "Output"an array of rank 2, 3 or 4
  • PoolingLayer can operate on arrays that contain "Varying" dimensions.
  • Possible explicit settings for the "Function" option include:
  • Maxthe maximum is used
    Meanthe mean value is used
    Totalthe sum of all values is used
  • When it cannot be inferred from other layers in a larger net, the option "Input"->{d1,…,dn} can be used to fix the input dimensions of PoolingLayer.
  • Given an input array of dimensions d1×…×di×…, the output array will be of dimensions ×…××…, where the channel dimension remains unchanged (i.e. d1=) and the sizes of spatial dimensions are transformed according to , where / are the padding sizes at the beginning/end of the axis, is the kernel size, and is the stride size for each dimension.
  • Options[PoolingLayer] gives the list of default options to construct the layer. Options[PoolingLayer[…]] gives the list of default options to evaluate the layer on some data.
  • Information[PoolingLayer[…]] gives a report about the layer.
  • Information[PoolingLayer[…],prop] gives the value of the property prop of PoolingLayer[…]. Possible properties are the same as for NetGraph.

Examples

open allclose all

Basic Examples  (2)

Create a PoolingLayer with a kernel size of 5×5:

Create a one-dimensional PoolingLayer:

Apply the layer to an input matrix:

Scope  (4)

Create a one-dimensional PoolingLayer with a stride of 4 and apply it to an input:

Create an initialized three-dimensional PoolingLayer with kernel size 2:

Apply the layer to an input array:

Create a two-dimensional PoolingLayer with a non-symmetric stride and apply it to an input:

Create a two-dimensional PoolingLayer that takes an image and returns an image:

Apply the layer to an image:

The layer threads across a batch of examples:

Options  (6)

"Function"  (2)

Different aggregation functions can be specified for pooling:

Make a table of the effect of the different pooling functions on two test images:

Interleaving  (1)

Create a PoolingLayer with InterleavingFalse and one input channel:

Create a PoolingLayer with InterleavingTrue and one input channel:

PaddingSize  (3)

Create a two-dimensional PoolingLayer that pads the first dimension with 10 zeros on each side and the second dimension with 12 zeros on each side:

Create a two-dimensional PoolingLayer that pads the first dimension with 10 zeros at the beginning and 14 zeros at the end:

Use padding to make the output dimensions equivalent to the input dimensions:

Properties & Relations  (2)

The following function computes the size of the non-channel dimensions, given the input size and parameters:

The output size of an input of size {256,252}, a kernel size of 3, a stride of 2, and padding size of 2:

This agrees with defining a PoolingLayer with the same parameters:

Increasing the stride can decrease the evaluation time of PoolingLayer:

Wolfram Research (2016), PoolingLayer, Wolfram Language function, https://reference.wolfram.com/language/ref/PoolingLayer.html (updated 2021).

Text

Wolfram Research (2016), PoolingLayer, Wolfram Language function, https://reference.wolfram.com/language/ref/PoolingLayer.html (updated 2021).

CMS

Wolfram Language. 2016. "PoolingLayer." Wolfram Language & System Documentation Center. Wolfram Research. Last Modified 2021. https://reference.wolfram.com/language/ref/PoolingLayer.html.

APA

Wolfram Language. (2016). PoolingLayer. Wolfram Language & System Documentation Center. Retrieved from https://reference.wolfram.com/language/ref/PoolingLayer.html

BibTeX

@misc{reference.wolfram_2024_poolinglayer, author="Wolfram Research", title="{PoolingLayer}", year="2021", howpublished="\url{https://reference.wolfram.com/language/ref/PoolingLayer.html}", note=[Accessed: 21-November-2024 ]}

BibLaTeX

@online{reference.wolfram_2024_poolinglayer, organization={Wolfram Research}, title={PoolingLayer}, year={2021}, url={https://reference.wolfram.com/language/ref/PoolingLayer.html}, note=[Accessed: 21-November-2024 ]}