Malt: A Deep Learning Framework for Racket - Anurag Mendhekar and Daniel P Friedman - RacketCon 2023

We discuss the design of a deep learning toolkit, Malt, that has been built for Racket. Originally designed to support the pedagogy of The Little Learner-A Straight Line to Deep Learning, it is used to build deep neural networks with a minimum of fuss using tools like higher-order automatic differentiation and rank polymorphism. The natural, functional style of AI programming that Malt enables can be extended to much larger, practical applications. We present a roadmap for how we hope to achieve this so that it can become a stepping stone to allow Lisp/Scheme/Racket to reclaim the crown of being the language for Artificial Intelligence (perhaps!).
con.racket-lang.org

[@tokkabokka8449]

The book is great but it doesn't do justice to automatic differentiation at the end. There needs to be another book 'Little differentiator' ! :)

[@happyboy2323]

好久不见！

[@mikeiavelli]

Racket is my preferred language, but I don't understand the python-bashing here; there are worst languages out there. I think it's doing a disservice to the whole project if you alienate possible future users with juvenile comments like that (e.g. 28:05). You might get enthusiastic racket programmers to contribute to the project, but you also need people from the established ML community, most of which happen to work in the python ecosystem...

[@KManAbout]

And many who do not like Python!

[@HG77K]

I appreciate the pedagogical perspective but why use Racket as a vehicle? What audience are you targeting here?

[@PragyAgarwal]

"directly mirrors the underlying math" - more like "directly obfuscates the underlying math" Nothing says readability like prefix notation for subtraction, amirite!?

[@LewisCampbellTech]

It's very easy to read once you learn to read it. Do you think you came out of the womb knowing about infix operators?