Miguel González Duque
@miguelgondu.bsky.social
Mathematician interested in optimization, probabilistic modelling, and differential geometry.
miguelgondu.com/about
miguelgondu.com/about
One of the developers of AlphaFold 3 on removing equivariance and invariance from the model: "It was a freeing and beautiful experience". #NeurIPS2024
December 15, 2024 at 5:40 PM
One of the developers of AlphaFold 3 on removing equivariance and invariance from the model: "It was a freeing and beautiful experience". #NeurIPS2024
It was great to present our work at #NeurIPS2024 and receive such positive feedback!
TL;DR: we survey high-dimensional Bayesian optimization, and find that current benchmarking practices are lacking standardization. We then contribute two open source libraries for benchmarking. Check them out!
TL;DR: we survey high-dimensional Bayesian optimization, and find that current benchmarking practices are lacking standardization. We then contribute two open source libraries for benchmarking. Check them out!
December 13, 2024 at 5:22 AM
It was great to present our work at #NeurIPS2024 and receive such positive feedback!
TL;DR: we survey high-dimensional Bayesian optimization, and find that current benchmarking practices are lacking standardization. We then contribute two open source libraries for benchmarking. Check them out!
TL;DR: we survey high-dimensional Bayesian optimization, and find that current benchmarking practices are lacking standardization. We then contribute two open source libraries for benchmarking. Check them out!
I keep coming back to the idea of "model metrology" they propose in this paper: openreview.net/forum?id=btt...
Benchmarks as Microscopes: A Call for Model Metrology
Modern language models (LMs) pose a new challenge in capability assessment. Static benchmarks inevitably saturate without providing confidence in the deployment tolerances of LM-based systems, but...
openreview.net
December 2, 2024 at 2:00 PM
I keep coming back to the idea of "model metrology" they propose in this paper: openreview.net/forum?id=btt...
Have I already spread the gospel of uv here?
I feel the entire ML/Bio community should be migrating away from conda as soon as possible.
I can definitely recommend Astral's uv as an alternative: docs.astral.sh/uv/
I feel the entire ML/Bio community should be migrating away from conda as soon as possible.
I can definitely recommend Astral's uv as an alternative: docs.astral.sh/uv/
uv
docs.astral.sh
November 26, 2024 at 2:53 PM
Have I already spread the gospel of uv here?
I feel the entire ML/Bio community should be migrating away from conda as soon as possible.
I can definitely recommend Astral's uv as an alternative: docs.astral.sh/uv/
I feel the entire ML/Bio community should be migrating away from conda as soon as possible.
I can definitely recommend Astral's uv as an alternative: docs.astral.sh/uv/