parameters: Indices for biases and weights

In mgoulet847/tagi: Tractable Approximate Gaussian Inference in Neural Networks

Description

This function assigns indices for all weights and biases in the neural network.

Usage

parameters(NN)

Arguments

NN	List that contains the structure of the neural network

Details

Bias indices are assigned from 1 to the maximum number of biases for a given layer. Then, weight indices start where bias indices end plus one until all weights are assigned an indice. The number of weights for a given layer is the number of units in the previous layer times the number of units in the current one.

For example, if there are 10 units in the previous layer and 50 in the current one, then there would be 50 biases and 500 weights in the current layer. The bias indices would be from 1 to 50 and weight IDs from 51 to 550.

Value

NN with three new elements, each of size (number of layers -1) :

• Weight indices for each layer

• Bias indices for each layer

• Combined weight and bias indices for each layer

mgoulet847/tagi documentation built on Dec. 21, 2021, 5:10 p.m.