Jean Feydy

As far as I can tell, the problem comes from the fact that `gather_coo` and `gather_csr` are not properly linked to their backward operators, which are respectively `segment_coo` and `segment_csr`...

Wow, thanks a lot for your quick answers! Unfortunately, it looks like there's something strange happening in the TC compiler: **I could reproduce the bug on a completely different machine,...

Hi @ftynse , I submitted a bug report for the standard convolution. As for the log-convolution (i.e. this bug report), no choice of mapping options solves the problem: the output...

Hi @prigoyal , I'm not sure about this: the bug #179 is related to large convolution filters / loops, whereas this one happens whenever I combine a **non-trivial reduction** with...

Sure, here it is https://gist.github.com/jeanfeydy/155e0f2bd5bb9b1dd8a2a4f6e7109530 (Just a copy-paste of the OP; I didn't know that we could share scripts so easily...)

Hi @josephenguehard, @bcebere, Thanks for your PR and comment! I am working for a new GeomLoss release in ~2 weeks (I've been getting up to speed with reproducible containers this...

Hi @RaghuSpaceRajan , Indeed, thanks for your detailed report! This is visibly the result of an unfortunate copy-paste in a separate function... As you'll have remarked, a simple work-around is...

Hi @albeltra, @rsanchezgarc, Thanks for your interest in this library! As said by @alberta, 1D optimal sorting amounts to sorting / histogram equalization so there is probably no need to...

Hi @gabrielfougeron, Thanks for your interest in the library! To answer your questions: - The library supports float32 by default. It is by far the most relevant numerical type for...

You’re very welcome! If you see a strange behavior, please let me know about it with an issue: I don’t have the necessary bandwidth to fix them “on-the-fly”, but I...