fcCombinaisonDweightNodeAll: All possible combinations of products of first derivatives

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

Description

This function calculates mean of products of first derivatives wd*wd. Since both weight and node are iterated over all products, every products (weight times node) from current layer are considered which results in a (Bni x no x noni)-array. I.e. each dimension of the array represents a single product being multiplied to all other possible products from current layer. Order is as followed: w11d1, w12d2, w13d3, ..., w1nidni, w21d1, w22d2, ..., w2nidni, ... wno1d1, ..., wnonidni

Usage

fcCombinaisonDweightNodeAll(mpdi, mpdin, mpdiw, ni, no, B)

Arguments

mpdi	Mean vector of first derivative product wd of current layer
mpdin	Mean array of combination of products of first derivatives (iterations on nodes)
mpdiw	Mean array of combination of products of first derivatives (iterations on weights)
ni	Number of units in current layer
no	Number of units in next layer
B	Batch size

Value

Mean array of combination of products of first derivatives

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