Matt Haberland

Thanks @ev-br @j-bowhay!

@HerculesJack thought I'd check in on this PR. Would you like to finish it up? What do you need from us?

Thanks for the summary @HerculesJack. In that case, I'll probably record the feature idea somewhere and close the PR. @lzkelley I've thought about the idea of leaving KDE enhancements to...

Thanks again @HerculesJack. One last question: this PR adds methods for computing the pdf/logpdf of a marginal distribution of a KDE distribution. I'm thinking that rather than having a separate...

OK, I resolved merge conflicts and added a commit (https://github.com/scipy/scipy/pull/9932/commits/0157c13325d54e5935b2c122c13fa53e037b8497) that shows the concept of the new approach. The tests (including one showing that old and new approaches produce the...

[Email to mailing list](https://mail.python.org/archives/list/[email protected]/thread/VZWWQR45WJKLDQPXNOSE6Y2WMLLF4WAL/) sent 8/11. No response so far, but we have a contributor and two stats maintainers who think this is a good idea. Yes, the implementation is...

@siddhantwahal It sounds like the originally-proposed changes (e.g. `einsum` or `matmul` instead of `tensordot`) were already big improvement and relatively simple. Would you like to revert to those so we...

Perfect. If it sounds good to you, I'll: - check for performance improvement. Assuming I can reproduce it, I'll continue. - merge main into this branch - add a commit...

I can confirm the speedup here, but it doesn't look like the random samples are the same (or close). Are they supposed to be the _same_, or is this just...

Would you like me to merge in main / 7dc3157 and push back here, @siddhantwahal?