Hi all, I have a huge sparse matrix, it's 360K*210K. Right now I have all the nonzero elements and their indices, so I want to transform this information into a sparse matrix so that I could pass the sparse matrix into svm function. I was planning to use sparse_matrix<- as.matrix.csr(v,nrow=360,ncol=300k), where v would be a vector with a length = 360K*210K and with nonzero elements in their places. However I could not even create a vector of that size: vector <-mat.or.vec( 1.08e+11,1), Error in numeric(nr) : vector size specified is too large. So I thought to transform only part of the data at a time: so I tried smaller size vector. As I found out, apparently the longest length you can create through mat.or.vec( ) function is 2*10^9, but that means I will have to break the data into 54 parts and transform each of them into a long vector and then transform the vector into sparse matrix and then rbind the 54 matrices into one . But I was worried that even if I could transform that data into 54 smaller sparse matrix, I wouldn't be able to rbind them together because of its huge size. I don't want to go through the arduous process only to find that they can't be stacked together... So I would really love to know how to transform this information into a sparse matrix ? And even if it can be done, can svm function in package e1071 handle a sparse matrix of this size ? Thanks!! [[alternative HTML version deleted]]