R/rbmUnitFunctions.R

In darch: Package for Deep Architectures and Restricted Boltzmann Machines

# Copyright (C) 2013-2016 Martin Drees
# Copyright (C) 2015-2016 Johannes Rueckert
#
# This file is part of darch.
#
# darch is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# darch is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with darch. If not, see <http://www.gnu.org/licenses/>.

#' @include rbm.Class.R
NULL

#' Calculates the RBM neuron output with the sigmoid function
#' 
#' Calculates the RBM neuron output with the sigmoid function from input saved
#'in \code{data}. 
#' 
#' @details The return value is a list with the output of the sigmoid function 
#' as first entry and binary representation calculated through a comparison of 
#' the output with random numbers. The random numbers a generated with the 
#' function \code{\link{runif}}.
#' 
#' @param rbm An instance of the class \code{\linkS4class{RBM}}.
#' @param data A matrix with the data for the calculations.
#' @param biases The biases for the calculations.
#' @param weights The weight matrix for the calculations.
#' @param runParams Parameters which indicates the status of the training.
#' @param matMult Matrix multiplication function.
#' @param ... Additional parameters, not used.
#' @return The real value and binary activations for the units
#' @family RBM unit functions
#' @export
#' @keywords internal
sigmoidUnitRbm <- function(rbm, data, biases, weights, runParams,
  matMult = getParameter(".matMult", ...), ...)
{
  ret = list()

  numUnits <- ncol(biases)
  batchSize <- nrow(data)
  randomNums <- matrix(runif(batchSize*numUnits),batchSize,numUnits)  
  ret[[1]] <- (1/(1 + exp(matMult(-data, weights) - kronecker(matrix(1,batchSize,1),biases))))
  ret[[2]] <- (ret[[1]] > randomNums)*1
  ret
}

#' Calculates the neuron output with the hyperbolic tangent function
#' 
#' Calculates the neuron output with the hyperbolic tangent function from
#' input in \code{data}.
#' 
#' @details The return value is a list with the output of the hyperbolic
#' tangent function as first entry and binary (-1,1) representation calculated
#' through a comparison ofthe output with random numbers. The random numbers a
#' generated with the function \code{\link{runif}}.
#' 
#' @param rbm A instance of the class \code{\linkS4class{RBM}}.
#' @param data A matrix with the data for the calculations.
#' @param biases The biases for the calculations.
#' @param weights The weight matrix for the calculations.
#' @param runParams Parameters which indicates the status of the training.
#' @param matMult Matrix multiplication function.
#' @param ... Additional parameters, used for parameter resolution.
#' @return The real value and binary (-1,1) activations for the units.
#' @family RBM unit functions
#' @export
#' @keywords internal
tanhUnitRbm <- function(rbm, data, biases, weights, runParams,
  matMult = getParameter(".matMult", ...), ...)
{
  ret = list()
  
  numUnits <- ncol(biases)
  batchSize <- nrow(data)
  randomNums <- matrix(runif(batchSize * numUnits, min = -1, max = 1),
                       batchSize, numUnits)
  ret[[1]] <- tanh(matMult(data, weights) + kronecker(matrix(1, batchSize, 1), biases))
  #ret[[2]] <- (ret[[1]] > randomNums)*2-1
  ret[[2]] <- (ret[[1]] > randomNums)*1
  ret
}

#' Calculates the linear neuron output no transfer function
#' 
#' Calculates the linear neuron output with no transfer function from real
#' value input saved in \code{data}.
#' 
#' @details The return value is a list with the output of the neurons as first
#' entry and binary representation calculated through a comparison of the
#' output with random numbers. The random numbers a generated with the
#' function \code{\link{rnorm}}.
#' 
#' @param rbm A instance of the class \code{\linkS4class{RBM}}.
#' @param data A matrix with the data for the calculations.
#' @param biases The biases for the calculations.
#' @param weights The weight matrix for the calculations.
#' @param runParams Parameters which indicates the status of the training.
#' @param matMult Matrix multiplication function.
#' @param ... Additional parameters, used for paramete resolution.
#' @return The real value and binary activations for the units
#' @family RBM unit functions
#' @export
#' @keywords internal
linearUnitRbm <- function(rbm, data, biases, weights, runParams,
  matMult = getParameter(".matMult", ...), ...)
{
  ret = list()
  
  numUnits <- dim(biases)[2]
  batchSize <- nrow(data)
  randomNums <- matrix(rnorm(batchSize*numUnits),batchSize,numUnits)
  ret[[1]] <- matMult(data, weights) + kronecker(matrix(1,batchSize,1),biases)
  #ret[[2]] <- ret[[1]] + randomNums
  ret[[2]] <- (ret[[1]] > randomNums)*1.
  
  ret
}