spatialDropoutLayer

Spatial dropout layer

Since R2024a

Description

A spatial dropout layer randomly selects input channels with a given probability, and sets all its elements to zero during training.

Creation

Syntax

layer = spatialDropoutLayer

layer = spatialDropoutLayer(Name=Value)

Description

layer = spatialDropoutLayer creates a spatial dropout layer.

layer = spatialDropoutLayer(Name=Value) sets the optional Probability and Name properties using name-value arguments. For example, spatialDropoutLayer(Probability=0.25) creates a spatial dropout layer with dropout probability 0.25. spatialDropoutLayer(Name="spat1") creates a spatial dropout layer with name "spat1".

Properties

Dropout

Probability — Probability to drop out all elements of an input channel

0.5 (default) | nonnegative number less than 1

Probability for dropping out all input elements in a channel, specified as a nonnegative number less than 1.

For an input of size [sizeDim1,...,sizeCDim, sizeBDim,...,sizeDimN], where sizeCDim is the number of channels, and sizeBDim is the batch size, a dropout mask is created using rand([1,...,sizeCDim, sizeBDim,...,1])<Probability. The input to the rand function is a vector of ones except at the locations of the channel dimension and the batch dimension. At training time, the layer randomly sets all the input elements to zero according to the dropout mask. It then scales the remaining elements by 1/(1-Probability). This operation effectively improve generalization performance by preventing activations from becoming strongly correlated [1], and helps prevent the network from overfitting. A higher probability results in more elements being dropped during training. At prediction time, the output of the layer is equal to its input.

Example: 0.4

Layer

Name — Layer name

'' (default) | character vector | string scalar

Layer name, specified as a character vector or a string scalar. For Layer array input, the trainnet and dlnetwork functions automatically assign names to unnamed layers.

The SpatialDropoutLayer object stores this property as a character vector.

Data Types: char | string

NumInputs — Number of inputs

Read-only: 1 (default)

This property is read-only.

Number of inputs to the layer, stored as 1. This layer accepts a single input only.

Data Types: double

InputNames — Input names

Read-only: {'in'} (default)

This property is read-only.

Input names, stored as {'in'}. This layer accepts a single input only.

Data Types: cell

NumOutputs — Number of outputs

Read-only: 1 (default)

This property is read-only.

Number of outputs from the layer, stored as 1. This layer has a single output only.

Data Types: double

OutputNames — Output names

Read-only: {'out'} (default)

This property is read-only.

Output names, stored as {'out'}. This layer has a single output only.

Data Types: cell

Examples

Create Spatial Dropout Layer

Create a spatial dropout layer with name "spat_drop1" and dropout probability 0.25.

layer = spatialDropoutLayer(Name="spat_drop1",Probability=0.25)

layer = 
  SpatialDropoutLayer with properties:

           Name: 'spat_drop1'
    Probability: 0.2500

   Learnable Parameters
    No properties.

   State Parameters
    No properties.

  Show all properties

Include a spatial dropout layer in a Layer array.

layers = [ ...
    imageInputLayer([28 28 1])
    convolution2dLayer(5,20)
    reluLayer
    spatialDropoutLayer
    fullyConnectedLayer(10)
    softmaxLayer]

layers = 
  6×1 Layer array with layers:

     1   ''   Image Input       28×28×1 images with 'zerocenter' normalization
     2   ''   2-D Convolution   20 5×5 convolutions with stride [1  1] and padding [0  0  0  0]
     3   ''   ReLU              ReLU
     4   ''   Spatial Dropout   Spatial Dropout
     5   ''   Fully Connected   10 fully connected layer
     6   ''   Softmax           softmax

References

[1] Jonathan Tompson, Ross Goroshin, Arjun Jain, Yann LeCun, and Christoph Bregler. "Efficient Object Localization Using Convolution Networks." arXiv preprint arXiv:1411.4280v3 (2015)

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

GPU Code Generation
Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Version History

Introduced in R2024a