(not that there's anything wrong with that)

the website has a list: cleanlabelproject.org/product-cate... which is specific and helpful, though oddly it slightly differs from the brands in the report:

i cooked him

pretty pretty good

Twitter is still a better place to share ML papers than Bluesky:

Finally, we provide easy-to-use software that "just works". We show results on a wide variety of datasets, from single-cell data, to gene expression, and even to image color adaptation! 7/7

Then, we propose a general framework for solving it. Basically, we propose minimizing an expectation over the divergence between the conditional distribution between the different domains. This reduces the problem to learning a conditional generative model on your data. 6/7

This paper formalizes the problem and offers a concrete, flexible solution: openreview.net/forum?id=GSp... . We begin by formalizing this problem as a small modification of generalized label shift (which combines label shift and covariate shift). 5/7

Excited to share a new paper published in TMLR, "Towards Backwards-Compatible Data with Confounded Domain Adaptation", solving a key problem in using AI for biology. How can you combine datasets across settings, when "what you measure" and "how you measure it" are confounded? 1/7