represents a net layer that normalizes its input by averaging across neighboring input channels.

Details and Options

  • The following optional parameters can be included:
  • "Alpha"1.0scaling parameter
    "Beta"0.5power parameter
    "Bias"1.0bias parameter
    "ChannelWindowSize"2number of channels to average over
  • LocalResponseNormalizationLayer[][input] explicitly computes the output from applying the layer.
  • LocalResponseNormalizationLayer[][{input1,input2,}] explicitly computes outputs for each of the inputi.
  • LocalResponseNormalizationLayer is typically used inside NetChain, NetGraph, etc.
  • LocalResponseNormalizationLayer exposes the following ports for use in NetGraph etc.:
  • "Input"a rank-3 tensor
    "Output"a rank-3 tensor
  • When it cannot be inferred from other layers in a larger net, the option "Input"->{n1,n2,n3} can be used to fix the input dimensions of LocalResponseNormalizationLayer.
  • The output tensor is obtained via , where D_(ijk)=(alpha sum_( x = max(1, i-c))^( min(N, c+i))T_(xjk)^2+b)^(beta), Tijk is the input, is the "Alpha" parameter, is the "Beta" parameter, is the "ChannelWindowSize" and N is the number of channels in the input Tijk.


open allclose all

Basic Examples  (2)

Create a LocalResponseNormalizationLayer:

Click for copyable input

Create a LocalResponseNormalizationLayer with input dimensions specified:

Click for copyable input

Apply the layer to an input:

Click for copyable input

Scope  (1)

Possible Issues  (1)

See Also

ConvolutionLayer  PoolingLayer  BatchNormalizationLayer  NetChain  NetGraph  NetInitialize  NetTrain  NetExtract

Introduced in 2017