Layer: Fully connected layer for neural network

In gumeo/mnistr: Utility Functions for MNIST Dataset

Description

Layer encapsulates all the data needed for a fully connected layer.

Usage

Layer(activation, minibatchSize, sizeP, is_input = FALSE, is_output = FALSE,
  initPos, initScale)

Arguments

activation	function for neural network. Must be able to do elementwise calculations on a matrix and have the parameter deriv, which is a boolean indicating whether the derivative should be calculated
minibatchSize	Number of samples used for estimating the gradient
sizeP	vector of two values, number of inputs to this layer and number of outputs from this layer, ignoring bias values.
is_input	boolean indicating whether this is an input
is_output	boolean indicating whether this is output
initPos	boolean indicating whether weights should be initialized as positive
initScale	scalar for initialising wieghts, e.g. if it is 100, then the randomly sampled initial weights are scaled by 1/100.

Value

environment with the functions to set all the internal matricies and a function to forward propagate through the layer.

Examples

testLayer <- Layer(mnistr::reLU, 3, c(10,10),FALSE,FALSE,TRUE,1000)
testLayer$W$getter() # Check random weights

gumeo/mnistr documentation built on May 17, 2019, 9:27 a.m.