layer_instance_norm_plus_params_as_input: Instance Normalization Layer where the Parameters are passed...

In rdinnager/DDLL: What the Package Does (One Line, Title Case)

Description

This layer performs an instance normalization using parameters passed in the inputs. This is useful if you want to calculate the affine transformation parameters somewhere else in the network, so this should keep the backpropagation working fin.

Usage

layer_instance_norm_plus_params_as_input(object, output_params = TRUE,
  epsilon = 0.001, name = NULL, trainable = TRUE)

Arguments

object	Named list of input tensors. Should be named "images", "beta", and "gamma", for the image tensor, and the beta and gamma affine parameters respectively.
output_params	Should the affine parameters be outputted as well? Or just the normalized image? Default: TRUE.
epsilon	epsilon value. Small value to add for numerical stability.
name	Optional layer name
trainable	Is the layer trainable?

Value

A list of output tensors, or a single tensor.

rdinnager/DDLL documentation built on Nov. 5, 2019, 3:04 a.m.