[R] scaling of Amat still recommended for quadprog::solve.QP ?
Benjamin Tyner
btyner at gmail.com
Thu Sep 25 12:48:07 CEST 2014
Greetings,
I ran across this recommendation, to keep the norms of the columns of
the Amat on similar
scales,
https://stat.ethz.ch/pipermail/r-help/2007-September/141335.html
However, when I looked at the code, I noticed that the norms are already
being calculated:
c
c calculate the norm of each column of the A matrix
c
do 51 i=1,q
sum = 0.d0
do 52 j=1,n
sum = sum + amat(j,i)*amat(j,i)
52 continue
work(iwnbv+i) = sqrt(sum)
51 continue
nact = 0
iter(1) = 0
iter(2) = 0
50 continue
though I am not sure exactly how these norms get used subsequently. My
question is, is it no longer necessary to follow Berwin's recommendation
from 2007? Or are the norms being calculated for some other purpose, and
the recommendation still applies?
Regards
Ben
More information about the R-help
mailing list