"CgreedySearch.MBparams.MBnetwork.MBdataDE.MBgibbsPars" <-
function(params,graph,data,gibbsPars,pars,grid0,grid1=NULL,style=c("geneCV","metaCV","geneMAX","knowMAX","metaMAX","geneROC","corrPROF","likeliCV","priorENRICH"),bind=TRUE,quick=TRUE,dist=FALSE,quiet=FALSE,...)
{
# Compatibility checks
    isCompat <- compatObjects(params,graph,chkvals=FALSE)
    if(!length(pars) || !all(pars %in% 1:params@m))
	stop("Invalid pars specification")
    if(is.null(grid0))
	stop("grid0 must be non-null")
    if(is.null(grid1) || bind)
	grid1 <- grid0
    grid0 <- as.matrix(grid0)
    grid1 <- as.matrix(grid1)
    if((ncol(grid0) == 1) && (length(pars) > 1))
	grid0 <- matrix(rep(grid0,length(pars)),nrow=3)
    if((ncol(grid1) == 1) && (length(pars) > 1))
	grid1 <- matrix(rep(grid1,length(pars)),nrow=3)
    if(!all(c(nrow(grid0),nrow(grid1)) %in% 3) || !all(c(ncol(grid0),ncol(grid1)) %in% length(pars)))
	stop("grid0 and/or grid1 have improper dimensions")
    if(!is.numeric(grid0) || !is.numeric(grid1) || any(is.na(grid0)) || any(is.na(grid1)))
	stop("All elements in grid0 and grid1 must be defined")
    if(!all(grid0[3,] > 0) || !all(grid1[3,] > 0))
	stop("All step sizes must be > 0")
    NIter0 <- round((grid0[2,]-grid0[1,])/grid0[3,]+1)
    NIter1 <- round((grid1[2,]-grid1[1,])/grid1[3,]+1)
    if(any(NIter0 <= 0) || any(NIter1 <= 0))
	stop("Ranges (max-min) must all be >= 0 and if > 0, >= step size")
    style <- match.arg(style)

# Iterate over all pars (possibly more than once if !quick greedy search)
    orig.pars <- pars
    for(tmp in orig.pars) params <- setZero(x=params,pNum=tmp)
    if(!quick) meta.score.best <- rep(-1,length(orig.pars))
    for(iter in 1:length(orig.pars)) {
	if(!quick && !quiet) cat(paste("\nStarting iteration ",iter," of ",length(orig.pars)," in greedy search...\n",sep=""))

	##### 1 round of 'quick' greedy search
	pars.best <- matrix(nrow=2,ncol=length(pars))
	score.best <- rep(-1,length(pars))
	for(ind in 1:length(pars)) {
	    if(!quiet) cat(paste("\nLearning parameter ",pars[ind]," (",ind," of ",length(pars),")...\n",sep=""))
	    pars.best[,ind] <- c(grid0[1,ind],grid1[1,ind])
	    count <- 1
	    par0 <- grid0[1,ind]
	    for(i in 1:NIter0[ind]) {

		# If parameters are not bound
		if(!bind) par1 <- grid1[1,ind]
		for(j in 1:NIter1[ind]) {
		    if(bind) break
		    params <- setValue(x=params,pNum=pars[ind],c(par0,par1))
		    score.temp <- switch(style,
                                       geneCV = CcrossValid(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       metaCV = CcrossValidMeta(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,what=params@Names[pars[ind]],...), # Function not yet written
                                       geneMAX = CpermSigGene(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       knowMAX = CpermSigKnow(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       metaMAX = CpermSigMeta(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,what=params@Names[pars[ind]],...),
                                       geneROC = CrocAuc(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       corrPROF = CcorrProf(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       likeliCV = CcrossValidLik(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       priorENRICH = CpriorEnrich(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...))
		    if(dist) cat(paste(score.temp,"\t",sep=""))
		    if(!is.na(score.temp) && (score.temp > score.best[ind])) {
			score.best[ind] <- score.temp
			pars.best[,ind] <- c(par0,par1)
		    }
		    if(!quiet) cat(paste("\nDone with ",count," of ",NIter0[ind]*NIter1[ind],"\n",sep=""))
		    count <- count+1
		    par1 <- par1+grid1[3,ind]
		}

		# If parameters are bound
		if(bind) {
		    params <- setValue(x=params,pNum=pars[ind],par0)
		    score.temp <- switch(style,
                                       geneCV = CcrossValid(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       metaCV = CcrossValidMeta(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,what=params@Names[pars[ind]],...), # Function not yet written
                                       geneMAX = CpermSigGene(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       knowMAX = CpermSigKnow(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       metaMAX = CpermSigMeta(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,what=params@Names[pars[ind]],...),
                                       geneROC = CrocAuc(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       corrPROF = CcorrProf(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       likeliCV = CcrossValidLik(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...),
                                       priorENRICH = CpriorEnrich(params=params,graph=graph,data=data,gibbsPars=gibbsPars,quiet=quiet,...))
		    if(dist) cat(paste(score.temp,"\t",sep=""))
		    if(!is.na(score.temp) && (score.temp > score.best[ind])) {
			score.best[ind] <- score.temp
			pars.best[,ind] <- par0
		    }
		    if(!quiet) cat(paste("\nDone with ",count," of ",NIter0[ind],"\n",sep=""))
		    count <- count+1
		}

		par0 <- par0+grid0[3,ind]
		if(dist) cat("\n")
	    }
	    if(quick) params <- setValue(x=params,pNum=pars[ind],pars.best[,ind])
	    else params <- setZero(x=params,pNum=pars[ind])
	    if(!quiet) cat(paste("\nDone with parameter ",pars[ind],"\n",sep=""))
	}
	##### end 1 round

	if(quick) break
	max <- which.max(score.best)
	params <- setValue(x=params,pNum=pars[max],pars.best[,max])
	meta.score.best[which(orig.pars == pars[max])] <- score.best[max]
	pars <- pars[-max]
	NIter0 <- NIter0[-max]
	NIter1 <- NIter1[-max]
	grid0 <- grid0[,-max]
	grid1 <- grid1[,-max]
	if(!quiet) cat(paste("\nDone with iteration ",iter,"\n",sep=""))
    }
    if(!quick) score.best <- meta.score.best
    if(dist) cat(paste("\nBEST scores:\n",score.best,"\n",sep=""))

    return(params)
}