Hacker News Reader

Got a source?

This is a lecture series, not new research, but one of the main takeaways is if you want good performance on modern machines, you had better be using arrays: https://youtube.com/playlist?list=PLHxtyCq_WDLXFAEA-lYoRNQIe...

I'm not OP, but I'm guessing he could be referring to something like this talk? https://www.youtube.com/watch?v=LrVi9LHP8Bk

I'm honestly not sure otherwise.

I think OP is talking about ML and matrix (array) multiplication to learn the right weights in a neural network.