Angelo Cortez

@wdevazelhes Wow thank you so much! Got around .93 AUC on my synthetic dataset example. Someone in my group also created a Mahalanobis distance based classifier that is similar performance....

> > the current LMNN is even more strict than the newer PR (it needs at least n_neighbors + 1 elements in each class, not just 2), so the new...

> @angelotc I'm glad it worked quite well, and indeed, I guess negative pairs are very important to learn the decision frontier, interesting curve ! It's nice to see you...

Coincidentally, [this paper](https://arxiv.org/pdf/2101.03778v1.pdf) basically models what we are attempting to explore with the exception that it is in the materials field (not dialog systems though I do work with dialog...

Hello, I am trying this out on another dataset now for superconductors cuprate classification. Running into an issue where ind_a and ind_b arrays are getting corrupted. I believe this is...

> @angelotc, did you try the new commit from @johny-c on your example again ? It would be interesting to know if it works now Yes, the algorithm performs well...

Looking forward to this development if it is still a thing

> I think you are spot on. I got this message regarding this: > > > FutureWarning: Comparing Series of datetimes with 'datetime.date'. Currently, the > > 'datetime.date' is coerced...

Maybe RAG style implementation (easy) or fine-tune LLM example (hard).