Category theory for data science: cautious optimism

I’m cautiously optimistic about applications of category theory. I believe category theory has the potential to be useful, but I’m skeptical of most claims to have found useful applications. Category theory has found useful application, especially behind the scenes, but a lot of supposed applications remind me of a line from Colin McLarty:

[Jean-Pierre] Serre created a series of concise elegant tools which Grothendieck and coworkers simplified into thousands of pages of category theory.

To be fair, Grothendieck put these recast tools to good use, but most mere mortals would stand a better chance of being able to use Serre’s versions.

My interest in category theory waxes and wanes, and just as it was at its thinnest crescent phase I ran across CQL, categorical query language. I haven’t had time to look very far into it, but it seems promising. The site’s modest prose relative to the revolutionary rhetoric of some category enthusiasts makes me have more confidence that the authors may be on to something useful.

Related post: Categorical (Data Analysis) vs (Categorical Data) Analysis

One thought on “Category theory for data science: cautious optimism

  1. nice post and hilarious quote! thanks. I agree with you that the field is not very approachable but this is why we’ve committed to taking some parts of that field that are particularly useful and applying those, conexus CQL is one (we (https://statebox) are developing a Haskell version of CQL in collab with Ryan & David,, but there are many other candidate subjects, one of them, statebox; which relies heavily on monoidal categories and diagrammatic languages. There is also hope in the Applied Category Theory field, it started as a meme but we are now starting to see this first serious applications and interest from industry.


Comments are closed.