R/norm.fun.R

In gecko515/HEMDAG: Hierarchical Ensemble Methods for Directed Acyclic Graphs

##**********##
## MAX-NORM ##
##**********##
#' @title Max normalization
#' @description Function to normalize the scores of a flat scores matrix per class.
#' @details The scores of each class are normalized by dividing the score values for the maximum score of that class.
#' If the max score of a class is zero, no normalization is needed, otherwise \code{NaN} value will be printed as results of 0 out of 0 division.
#' @param S matrix with the raw non normalized scores. Rows are examples and columns are classes.
#' @return A score matrix with the same dimensions of \code{S}, but with scores max/normalized separately for each class.
#' @export
#' @examples
#' data(scores);
#' maxnorm <- normalize.max(S);
normalize.max <- function(S){
    classes <- colnames(S);
    maximum <- apply(S,2,max);
    for(class in classes){
        if(maximum[class] != 0){
            S[,class] <- S[,class]/maximum[class];
        }
    }
    return(S);
}

##********************##
## FLAT NORMALIZATION ##
##********************##
#' @title Scores Normalization Function
#' @description Functions to normalize a flat scores matrix w.r.t. max normalization (MaxNorm) or quantile normalization (Qnorm) 
#' @details To apply the quantile normalization the \pkg{preprocessCore} package must be installed.
#' @param norm.type can be one of the following two values:
#' \itemize{
#' \item MaxNorm (\code{def.}): each score is divided w.r.t. the max of each class;
#' \item Qnorm: a quantile normalization is applied. Library preprocessCore is used;
#' }
#' @param S a named flat scores matrix with examples on rows and classes on columns.
#' @return the matrix of the scores flat normalized w.r.t. MaxNorm or Qnorm.
#' @export
#' @examples
#' data(scores);
#' norm.types <- c("MaxNorm","Qnorm");
#' for(norm.type in norm.types){
#'     scores.normalization(norm.type=norm.type, S=S)
#' }
scores.normalization <- function(norm.type="MaxNorm", S){
    if(norm.type=="MaxNorm"){
        ## Max Normalization
        S <- normalize.max(S);        
    }else if(norm.type=="Qnorm"){
        ## Quantile Normalization 
        ## NOTE: normalize.quantiles function returns a unnamed matrix. colnames are essential for hier.corr..
        S.norm <- normalize.quantiles(S);
        dimnames(S.norm) <- list(rownames(S),colnames(S));
        S <- S.norm;
        rm(S.norm);    
    }else{
        stop("scores.normalization: the chosen normalization method is not among those available or it has been misspelled");
    }
    return(S);
}


#' @title Flat scores normalization
#' @description High level functions to normalize a flat scores matrix w.r.t. max normalization (MaxNorm) or quantile normalization (Qnorm). 
#' @details To apply the quantile normalization the \pkg{preprocessCore} package must be installed.
#' @param norm.type can be one of the following two values:
#' \itemize{
#' \item MaxNorm (\code{def.}): each score is divided w.r.t. the max of each class;
#' \item Qnorm: a quantile normalization is applied. Library preprocessCore is used.
#' }
#' @param flat.file name of the flat scores matrix (without rda extension).
#' @param flat.dir relative path to folder where flat scores matrix is stored .
#' @param flat.norm.dir the directory where the normalized flat scores matrix must be stored.
#' @return the matrix of the scores flat normalized w.r.t. MaxNorm or Qnorm in \code{flat.norm.dir}.
#' @export
#' @examples
#' data(scores);
#' tmpdir <- paste0(tempdir(),"/");
#' save(S, file=paste0(tmpdir,"scores.rda"));
#' flat.dir <- flat.norm.dir <- tmpdir;
#' norm.types <- c("MaxNorm","Qnorm");
#' for(norm.type in norm.types){
#'     Do.flat.scores.normalization(norm.type=norm.type, flat.file="scores", 
#'     flat.dir=flat.dir, flat.norm.dir=flat.norm.dir);
#' }
Do.flat.scores.normalization <- function(norm.type="MaxNorm", flat.file=flat.file, 
    flat.dir=flat.dir, flat.norm.dir=flat.norm.dir){
    
    ## loading flat scores matrix 
    flat.path <- paste0(flat.dir, flat.file,".rda");
    S <- get(load(flat.path));

    ## scores normalization
    S <- scores.normalization(norm.type=norm.type, S);
    ## Storing results
    save(S, file=paste0(flat.norm.dir, norm.type, ".", flat.file, ".rda"));
}