R/svd.als.R

In softImpute: Matrix Completion via Iterative Soft-Thresholded SVD

svd.als=function(x, rank.max=2, lambda=0,thresh = 1e-05, maxit=100,trace.it=FALSE,warm.start=NULL,final.svd=TRUE){
  if(rank.max>(rmax<-min(dim(x)))){
    rank.max=rmax
    warning(paste("rank.max should not exceed min(dim(x)); changed to ",rmax))
  }
  ismiss=is.na(x)
  if(any(ismiss))stop("NAs in x; use softImpute instead")
  this.call=match.call()
  out=simpute.als(x,J=rank.max,thresh,lambda,maxit,trace.it,warm.start,final.svd)
  attr(out,"call")=this.call
  attr(out,"lambda")=lambda
  out
}

svd.als.sparse=function(x, rank.max=2, lambda=0,thresh = 1e-05, maxit=100,trace.it=FALSE,warm.start=NULL,final.svd=TRUE){
   this.call=match.call()
  out=Ssvd.als(x,J=rank.max,thresh,lambda,maxit,trace.it,warm.start,final.svd)
  attr(out,"call")=this.call
  attr(out,"lambda")=lambda
  out
}

  setGeneric("svd.als",svd.als)
  setMethod("svd.als","sparseMatrix",svd.als.sparse)
  setMethod("svd.als","SparseplusLowRank",svd.als.sparse)