R/ARTMAP_Add_New_Category.r

In gbaquer3/fuzzyARTMAP: Implementation of a fuzzy ARTMAP neural network

#' Adds a new category to the given weight matrix and map field.
#' 
#' This function returns a new weight matrix which is identical to the
#' given weight matrix except that it contains one more category which
#' is initialized to all 1's. It also returns a new map field matrix
#' which is identical to the given map field matrix except that it contains
#' one more element initialized to 0.
#' @title ARTMAP_Add_New_Category
#' @param weight Matrix of size NumFeatures-by-NumCategories which holds the weights of the network.
#' @param mapField Matrix of size 1-by-NumCategories which holds the associations between categories and classes.
#' @return Pair of values [RESIZEDWEIGHT, RESIZEDMAPFIELD]. The RESIZEDWEIGHT is a matrix of size NumFeatures-by-NumCategories+1
#' which holds the weights of the old matrix plus a new category of all
#' values of 1. The RESIZEDMAPFIELD is a matrix of size 1-by-NumCategories+1
#' which holds the given map field matrix except that it contains
#' one more element initialized to 0.
#' 
ARTMAP_Add_New_Category=function(weight, mapField)
{
  # Make sure that the user specified the weight and map field matrices.
  if(nargs() != 2)
  {
    stop('You must specify the weight and map field parameters.');    
  }
  
  # Create the return weight matrix with the right dimensions.
  numFeatures=ncol(weight);
  numCategories= nrow(weight);
  
  newCategory = matrix(1,1,numFeatures);
  resizedWeights = rbind(weight,newCategory);
  
  # Create the return map field matrix with the right dimensions.
  resizedMapField = rbind(mapField, 0); #Assuming matrix and not concatenation result (c())
  
  
  return (list(resizedWeights,resizedMapField));
}