DenseOrderedBias: Create a Dense layer with ordered biases

In enweg/quickbnnr: Quick way to implement Bayesian Neural Networks in R and estimating these using Julia's Turing

Create a Dense layer with ordered biases

Description

Priors for weights are standard normal. The ordering of biases is implemented by putting a standard normal prior on the first bias, and a truncated standard normal prior (0, inf) on the other biases. The actual biases are then the cumulative sum of the biases.

Usage

DenseOrderedBias(
  in_size,
  out_size,
  activation = c("identity", "sigmoid", "tanh", "relu")
)

Arguments

in_size	input size
out_size	output size
activation	activation function

enweg/quickbnnr documentation built on April 15, 2022, 3:29 a.m.