You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current pseudoinverse (hard-coded) doesn't always converge because numpy uses an iterative(!) algorithm for SVD instead of a true factorization. The best solution I think is to default to the least-squares generalized inverse and allow users to use the SVD-based inverse if and only if they wish to. This also lets us eliminate the home-build pseudoinverse from the math library for Gopt.
@PaulWAyers Got a little busy this week. But I will take some this time weekend to check the SVD-based inverse function in GOpt. Will let you know the progress
The current pseudoinverse (hard-coded) doesn't always converge because numpy uses an iterative(!) algorithm for SVD instead of a true factorization. The best solution I think is to default to the least-squares generalized inverse and allow users to use the SVD-based inverse if and only if they wish to. This also lets us eliminate the home-build pseudoinverse from the math library for
Gopt
.This is what was done in the
Procrustes
package.https://github.com/theochem/procrustes/blob/ad70b01578152d4f3c7570c065c7e2b1e8ac553b/procrustes/generic.py#L88
@tczorro do you want to take a crack at this, or do you want me to do it?
The text was updated successfully, but these errors were encountered: