Erik Welch

Relatedly, the workaround I posted does not work reliably in JupyterLab (version 3.0.16). Here is a reproducer: ```python import string import threading from ipywidgets import Output from IPython.display import display...

I have begun [`dask-patternsearch`](https://github.com/eriknw/dask-patternsearch) project that may be useful here. You can think of it as a line-search algorithm generalized to multiple dimensions without using derivatives. Its distinguishing feature is...

Thanks for the suggestion @amueller and letting me know they have a distributed implementation. I plan to perform comparative benchmarks against a couple notable packages and will definitely include `spearmint`....

And I suppose the current best way to do this is to make a matrix with a single element on a diagonal and do a matrix multiply from the left...

btw, I found a different implementation, so I would not use these if available atm. Still, these _may_ be nice (but def. not necessary), especially if they provide a performance...

btw, I'm playing around with implementing square clustering coefficient, and one of my variants could use COLNE and ROWNE IndexUnaryOp. Do you know if square clustering coefficient is implemented anywhere...

Thanks for pointing out LAGraph's local clustering coefficient. I think I implemented this for undirected and directed, unweighted and weighted graphs. I should compare (and benchmark) against LAGraph. I use...

COLEQ, ROWEQ, COLNE, and ROWNE seem like the most natural to add. _If_ ROWCOLEQ and/or ROWCOLNE were added (up to you), yeah, I would have them be: - ROWCOLEQ: `(i...

Thanks. I understand this is the case for generic binary operators. If the internal data formats are sorted, I would expect there to be efficient *specialized* implementations for FIRST and...

fwiw, I can't think of a meaningful example *when restricted to the current spec* where a Monoid can't be used instead of a BinaryOp. So, I think it's fine that...