I have read a substantial part of Francois Chollet’s book Deep Learning with Python, here:
Honestly, you won’t find much new material if you have already followed both part 1 and 2 of our beloved course, but I found it a very useful read to refresh and revise concepts and ideas. It is very applied and code-rich, obviously on Francois’ Keras…
It is currently in “early access” mode, the hardcopy is expected for this fall I think.
Anyone else has read it too?
ps: I used the discount code “deeplearning” (found it on google…) and got a 42% discount, so in total it costs ~23$.
I didn’t know there was a discount, but I bought it anyway. So far, it is an excellent book. He really explains the concepts very well. Highly recommended.
I read the book last week, and can say this is an excellent companion text to fast.ai; it covers pretty much exactly the same material as jeremy’s courses, and can serve as a reference and handy collection of code snippets.
Some things not contained in the book: Wasserstein GAN, collaborative filtering, and pytorch (for obvious reasons).
I read it as well and thought it was great. If you told me it was written specifically for fast.ai I would believe you. Some cool tricks that it mentioned that weren’t in part 1:
Callbacks- things like automatically reducing the lr on plateau solve a lot of problems. Also access to tensorboard, early stopping, and automatic is really great.
GANs- were implemented in keras too, which was cool because most tutorials I see are in a more granular language.
Different Tools- hyperopt and arxiv-sanity are a few that come to mind
