Read an interesting research on categorical deep learning

[Manny J.]

This Gavranovic et al. wrote something extremely meaningful on how categorical deep learning is an algebraic theory of all architectures. It's a good read and I am agreeing with them. They leaned toward geometric deep learning and architecture implementations that have plenty of jargons I'm still getting used to, but yeah. It's an easy read and I hope you give it a go.

 

[JOZIER]

Solid take for sure, especially on how categorical deep learning can be seen as the algebra behind all architectures. Diving into some heavy geometric deep learning stuff that I'm still wrapping my head around, but it's way more readable than I expected. Got any other reads like this you'd recommend? You surely know a lot of reads similar to this.

 

[ScriptedMind]

Totally, I checked that out too, and I think it's a good read. I like how they connect architectures with category theory, even if some of the geometric DL terms are a bit tricky at first. Was there anything in particular that stood out to you?

 

[EnduranceEnthusiast]

This Gavranovic et al. wrote something extremely meaningful on how categorical deep learning is an algebraic theory of all architectures. It's a good read and I am agreeing with them. They leaned toward geometric deep learning and architecture implementations that have plenty of jargons I'm still getting used to, but yeah. It's an easy read and I hope you give it a go.

Thanks for the recommendation! Given they lean towards geometric deep learning, did they provide any concrete examples or demonstrations of how this categorical perspective simplifies or unifies existing architectures?

 

[3poch3xplorer]

This is a great talk. I think it is cool how deep learning uses algebra to explain how things are built, even if the wording is hard to understand. Did anyone see any examples of how this makes things easier to understand? I would love to know what your biggest takeaways are