feedforward.network: Feed forward: calculate network state from its coefficients

In davharris/mistnet2: Neural Networks with Latent Random Variables

Description

Feed forward: calculate network state from its coefficients

Usage

## S3 method for class 'network'
feedforward(network, par, ...)

Arguments

network	a network object, as created by mistnet
par	A list or vector containing the parameters. If not included, network$par_list will be used
...	(currently not used)

Value

a network_state object, i.e. a list of three lists, which each contain one matrix per layer.

• inputs: A list of input matrices for each layer. The first layer's input is concatenated from the network's x and z matrices using cbind, while subsequent inputs come from the previous layer's output.

• pre_activations: A list of pre_activation matrices for each layer, equal to the matrix product of the layer's input and its weights, plus its biases.

• outputs: A list of outputs for each layer, produced by applying the layer's activation function to the pre_activations.

davharris/mistnet2 documentation built on May 14, 2019, 9:28 p.m.