nnlayer_norm: Create response normalization layer.

In andrie/mxNeuralNetExtra: Helper Functions to Make it Easier to Use neuralNet in the MicrosoftML Package

Description

Create response normalization layer.

Usage

nnlayer_norm(layer, kernelshape, inputshape, name, inputname, stride, lowerpad,
  alpha = 1e-04, beta = 0.75, offset = 1, AvgOverFullKernel = TRUE,
  activation = NA)

Arguments

layer	A layer object, e.g. using nnlayer_input, or NULL
kernelshape	Numeric vector describing the number of inputs in each dimension, e.g. c(3, 15, 15)
inputshape	Numeric vector describing the number of inputs in each dimension, e.g. c(3, 15, 15). If layer is specified, this can be NULL.
name	Name of the layer
inputname	Name of the preceding layer. If layer is specified, this can be NULL.
stride	Numeric vector describing the number of inputs in each dimension, e.g. c(3, 15, 15).
lowerpad	Numeric vector describing the number of inputs in each dimension, e.g. c(3, 15, 15).
alpha	Alpha
beta	Beta
offset	Offset
AvgOverFullKernel	Logical. Take average over full kernel.
activation	Activation function, e.g. rlinear

See Also

Other layer.definition.functions: nnlayer_conv, nnlayer_full, nnlayer_input, nnlayer_output, nnlayer_pool

andrie/mxNeuralNetExtra documentation built on May 10, 2019, 11:19 a.m.