fcHiddenStateBackwardPass: Backpropagation (states' deltas) for fully connected layers...

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

Description

This function calculates units' deltas at a given layer when using more than one observation at the time.

Usage

fcHiddenStateBackwardPass(Sz, Sxs, J, mw, deltaM, deltaS, ni, no, B, rB)

Arguments

Sz	Covariance of units from current layer
Sxs	Null by default (not used yet)
J	Jacobian of current layer
mw	Mean vector of weights for the current layer
deltaM	Delta of mean vector of the next layer units given y μ_{Z}|y
deltaS	Delta of covariance matrix of the next layer units given y Σ_{Z}|y
ni	Number of units in current layer
no	Number of units in next layer
B	Batch size
rB	Number of times batch size is repeated

Value

• Delta of mean vector of current layer units given y μ_{Z}|y

• Delta of covariance matrix of current layer units given y Σ_{Z}|y

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