Deep Learning with Jax

[teleforce]

https://www.manning.com/books/deep-learning-with-jax

[xmaayy]

More of a meta-question; I'd love to know if people do self-directed learning more frequently, or if they follow through full (or partial) textbooks to learn concepts they're interested in. I have a hard time going through full books, but I have no doubt this leads to gaps in knowledge that someone following a textbook would not have.

[pthreads → xmaayy]

I do both. In my case textbooks are to fill theoretical needs/gaps. I go through full books in some cases and in others it is not feasible based on various factors (difficulty of subject, need, content, 500+ pages etc.) I find that some of the newer textbooks give a helpful flowchart of chapters so one can choose their "adventure" based on interest or the curriculum needs (of an instructor).

On occasions I abandon reading after a chapter or two if I don't think it serves my needs or the subject matter is too difficult for me. In the latter case I may try to look for an alternative or come back to it once I have gained pre-requisite knowledge.

[cjbgkagh → xmaayy]

I look for code-available seminal papers so I can read the code as I understand code better than written text or mathematical expressions.

In order to understand it I'll try extend it, repurpose it, 'use it in anger', or re-write to a new framework or language.

[panarky]

Jupyter notebooks for every chapter available here: https://github.com/che-shr-cat/JAX-in-Action

[Scene_Cast2]

I've never used Jax, but it keeps popping up as an interesting library. How does it fit in and compare to pytorch, numba, and numpy?

[eyeris → Scene_Cast2]

Should be lower level than pytorch. Not sure about numba, but it should be pretty similar to numpy.

[slt2021]

i bought early preview of this book like a year ago (through https://www.manning.com/meap-program), only when the first two chapters were ready.

waiting for a printed copy to arrive. this is really good book

[xrd → slt2021]

Can you say more about why, specifically?

[slt2021 → xrd]

i was not familiar with JAX, so it was a good resource for me into the basics of the language, its advantages and use cases.

I also loved the notebooks that come with the book and that every big idea in the book is explained with the code example

[che_shr_cat → slt2021]

Thank you for the feedback! I'm the author of the book :)

[xrd → slt2021]

You had me at notebooks! Thanks so much.

[jszymborski]

Looks super interesting :) I was considering picking up a printed copy, but shipping to Montreal, Canada is 31 USD, pretty much the price of the book.

[sudosysgen → jszymborski]

If you're affiliated with Mila, it may be worth it to ask for the book to be bought for the shared library - I'm sure many people would peruse it :)

[jszymborski → sudosysgen]

That makes a lot of sense! Might just order it through a local bookseller though.