New paper: Expressing NumPy Broadcasting via Verb Rank in J

[Marcin Żołek]

A new paper on J has just been published on arXiv: Expressing NumPy Broadcasting via Verb Rank in J.

Have a look! The paper shows how the two approaches to writing loopless code correspond, with a detailed explanation of rank in J, all using a new way of presenting J code in articles.

Cheers,

Marcin

[Raul Miller]

Hmm.... first thought on looking at this - what would be an
interesting use case for Broadcastly?

Thanks,

--
Raul

> To unsubscribe from this group and stop receiving emails from it, send an email to forum+un...@jsoftware.com.

[Marcin Żołek]

Actually the adverb Broadcastly along with the api/python3 addon is mostly for Python/NumPy users who want to extend their programs with fully fledged array programming. They can do this by writing J scripts that are executed from Python with data passed from Python to J and the results returned from J to Python. Broadcastly allows NumPy arrays with their broadcast dimensions to be used directly in J programs.

The main point of the paper is to show that translating broadcasting (core feature of NumPy) to J is straightforward. At the same time, I also tried to explain verb rank in J in a simple and clear way. The definition is used in the proofs of correctness of the main ideas.

Marcin

[Joe Hohertz]

This comes up a lot in LLM related areas. I look forward to giving it a read.