I'm playing around with matrix inversion schemes for my own edification. I figured I would give Newtons's method, which sure looks straightforward enough, a try. My calculation is going unstable no matter what I do. I am working in Matlab and doing straightforward matrix multiplies:
for i=1:its
matinv = 2*matinv - matinv*mat*matinv
end
Am I doing something wrong, like, for instance, should I be updating one row/column at a time or something? I haven't been able to find any details of that sort. I know that a bad starting point can cause trouble, but I have started at so many points now that I think it must be something more fundamental.