layer_glu: Gated-Linear-Unit keras lambda layer

In ifrit98/layerR:

Description

Computes gated temporal convolutions as in: 'hl(X) = (X∗W+b) ⊗ σ(X∗V+c)“, where 'X' is a data matrix, 'W' is a weight kernel, and 'b','c' are biases. '⊗' is the element-wise multiplication operator

Usage

layer_glu(
  object,
  filters,
  kernel_size,
  kernel_initializer = "glorot_normal",
  kernel_regularizer = NULL,
  bias_initializer = "zeros",
  bias_regularizer = NULL,
  name = NULL,
  trainable = TRUE
)

References

https://arxiv.org/abs/1612.08083

ifrit98/layerR documentation built on March 2, 2020, 8:11 a.m.