Seeking Guidance and Collaboration: Applied Category Theory Research in AI

Hello everyone,

I hope you’re all doing well! I’m a master’s student in mathematics with experience in data engineering, data science, and artificial intelligence. I’ve developed a keen interest in Applied Category Theory, particularly its intersection with AI.

Recently, I’ve been focusing my readings on the Category for AI school and, in particular, the work of Bruno Gavranović. I’m now looking to get involved in a research project that combines applied category theory with a computational component, where algorithms are implemented and tested.

A good example of the type of project I’m seeking is the Adjoint School project, “Compositional Generalization in Reinforcement Learning” by Georgios Bakirtzis. In the project description, they mention:

“This project will tackle the problem of compositional generalization in reinforcement learning in a category-theoretic computational framework in Julia. Expected outcomes of this project are category theory-derived algorithms and concrete experiments.”

This approach, blending theoretical foundations with practical experimentation, aligns closely with what I’m looking for.

I’m reaching out to explore collaboration opportunities, potential projects, or simply to gather advice on how to proceed. Specifically, I would appreciate:

  • Suggestions for ongoing or new projects that involve both theoretical and computational components in applied category theory and AI.
  • Recommendations for resources, tools, or techniques that would help bridge theory and implementation.
  • Advice on building collaborations with researchers or groups working on similar problems.

I’m eager to contribute and collaborate with like-minded individuals or groups in this community. If you have any ideas, opportunities, or guidance, I would be incredibly grateful!

Thank you for your time, and I look forward to hearing your thoughts.

Best regards,
Pierre R

Hi Pierre,

Thanks for reaching out! I have to say that I’m very much not up to date with the AI/ACT stuff, but I would suggest the following procedure for getting involved.

  1. If you haven’t already, join the category-theory zulip and ask around there.
  2. If you have a coding background, try just taking some papers and implementing them! You will quickly learn that ACT papers leave a lot underspecified for efficient, ergonomic implementation, so you may even run into some novel research this way.
  3. Take public notes on things that you read. Your perspective as a beginner is super important, and there are a lot of things where more beginner-oriented tutorials/notes would be super useful for the next person trying to learn this stuff. If you want to make notes in the localcharts forest, I’d be happy to give you push access.

And whatever you end up doing, please drop a link to it on localcharts!

-Owen

1 Like

Hi Owen,

Thank you for your on-point advice! I just posted on the Category Theory Zulip.

In fact, while looking at older posts, I found the YouTube video Emilio Minichiello — Decision Problems on Graphs with Sheaves a good start to explore.

I will definitely start sharing my journey here!

Thank you for your time.

  • P

Hi Pierre,

Have you seen Bartosz Milewski’s Haskell implementation yet? (paper: Publications/NeuralLens.pdf at master · BartoszMilewski/Publications · GitHub, talk: https://www.youtube.com/watch?v=Ri_oC2gf-aY).

You might get inspiration from @tsmithe’s dissertation and other work (his website here).

Also, there’s some foundational computation-related stuff in Bob Walter’s book, Categories and Computer Science.

Best,
JR

1 Like

Thank you, JR! Pretty interesting content.

1 Like