Category Theory
Zulip Server
Archive

You're reading the public-facing archive of the Category Theory Zulip server.
To join the server you need an invite. Anybody can get an invite by contacting Matteo Capucci at name dot surname at gmail dot com.
For all things related to this archive refer to the same person.

Stream: community: our work

Topic: Georgios Bakirtzis

Georgios Bakirtzis (Sep 11 2024 at 09:31):

I wanted to give a bit of an update on my work that intersects with this community.

• Our paper Categorical semantics of compositional reinforcement learning has finally been accepted at the Journal of Machine Learning Research with minor revisions. However, the reviewers had some really good feedback that was more expansive that my coauthors and I are addressing in the meantime. There will be an update to the arXiv as soon as we finish those.
• Our simulated results from the theory of the above paper, Reduce, Reuse, Recycle: Categories for Compositional Reinforcement Learning, got accepted in the 2024 European Conference on Artificial Intellince (ECAI).

We are now working on an extension trying to understand the concept of generalization in reinforcement learning from a categorical lens, but since I recently moved to Catalunya for a job at Universitat Politècnica de Catalunya - BarcelonaTech (UPC) as faculty, work has been rather slow. I am open to sharing more about the idea if people are interested however (it is in very draft form).

And a sidenote, I hope to make it to ACT next year, the past three years have been difficult with respect to finding a job and moving a lot (part of academic life) so it made it hard for me to stay on top of what this community is doing. If some of this work intersects with something I have not seen/read (especially in case it is your own work!) please put me back in the loop. As faculty I hope that I will find students (and money for those students) to continue working on this field. In particular, @Valeria de Paiva had some good ideas for using Dialectica categories for modeling multi-agent systems that I didn't follow up as much as I hoped and want to revisit. I also have some work on using pullbacks to come up with a more direct algorithm for physical contracts in control that we are now rewriting for submission. Just to not give the impression I am only working on reinforcement learning these days :)

Valeria de Paiva (Sep 11 2024 at 12:51):

Congrats on getting your academic job at UPC Barcelona, @Georgios Bakirtzis ! Splendid news! I have not done anything about multiagents and dialectica constructions. yet. Still on my TODO list. so we should chat sometime.