Gram Schmidt with R

Question

Here is a MATLAB code for performing Gram Schmidt in page 1 http://web.mit.edu/18.06/www/Essays/gramschmidtmat.pdf

I am trying for hours and hours to perform this wi

Answer 1

Here a version very similar to yours but without the use of the extra variabale v. I use directly the Q matrix. So no need to use drop. Of course since you have j-1 in the index you need to add the condition j>1.

f=function(x){
  m <- nrow(x)
  n <- ncol(x)
  Q <- matrix(0, m, n)
  R <- matrix(0, n, n)
  for (j in 1:n) {
    Q[, j] <- x[, j]
    if (j > 1) {
      for (i in 1:(j - 1)) {
        R[i, j] <- t(Q[, i]) %*% Q[, j]
        Q[, j] <- Q[, j] - R[i, j] * Q[, i]
      }
    }
    R[j, j] <- max(svd(Q[, j])$d)
    Q[, j] <- Q[, j]/R[j, j]
  }
  return(list(Q = Q, R = R))
}

EDIT add some benchmarking:

To get some real case I use the Hilbert matrix from the Matrix package.

library(microbenchmark)
library(Matrix)
A <- as.matrix(Hilbert(100))
microbenchmark(grahm_schimdtR(A),
               grahm_schimdtCpp(A),times = 100L)

Unit: milliseconds
expr       min         lq     median        uq        max neval
grahm_schimdtR(A) 330.77424 335.648063 337.443273 343.72888 601.793201   100
grahm_schimdtCpp(A)   1.45445   1.510768   1.615255   1.66816   2.062018   100

As expected CPP solution is really fster.