Evgeni Burovski

There are no known issues in scikit-learn, at least according to https://github.com/numpy/numpy/pull/30411#issuecomment-3636477320 If I missed something in scikit-learn, could you point me to it? The `scikit-image` issue does not exist...

cross-ref https://github.com/scipy/scipy/issues/21279 -- I wonder if this fix also fixes that other issue.

Close/reopen + merge main to flush the CI.

Looked at this again. I hope it's not too much to ask @DWesl to check if this fixes https://github.com/scipy/scipy/issues/22831 and https://github.com/scipy/scipy/issues/21279, too? Both seem to have MWEs. If these can...

scipy.interpolate README file can be just removed. Fitpack is not really vendored, what we carry has been heavily modified w.r.t. the netlib version over time. And the modernized implementation, who...

I believe the algorithmic content of this PR is a faithful rewrite of the FITPACK logic behind `RectBivariateSpline`. The knot selection logic is the same, and the linear algebra is...

The _further memory and performance improvements_ bullet looks worth investigating: https://github.com/scipy/scipy/pull/23547#issuecomment-3578205867 reported slowdowns of up to 25%, which may or may not be noise (on a different machine, I saw...

Several notes: API: - `solve` should be fully backwards compatible. Some kwargs could use a change, cf gh-23072, but these changes are not used here. - lapack dtype conversion is...

> could we just stick with inv first? ... I think it's a bit too much to handle all at once `inv` is in for a while now, and there...

Okay then, let's get this in and track further work in https://github.com/scipy/scipy/issues/23141.