The same Bishop who wrote PRML? Nice! Looking through it does seem to be very up to date, although I wish there was a little bit more on topics like geometric deep learning and flow matching. Of course if every niche had a more thorough treatment this book would be impossibly long but more than a couple paragraphs would have been cool.
where does this fit in? I am reading Intro to statistical learning, so wondering if this should come after it or I can directly jump to this when I reach the unsupervised ML techniques.
Jurafsky's 3rd edition draft of his NLP book and Simon Prince's DL book both have chapters on transformers, and the latter also on diffusion. Both have official free pdf versions.
Murphy's book is about probabilistic/Bayesian machine learning whereas this, according to the preface is 'almost entirely non-Bayesian'. Murphy's book is also a lot more general, this one is about the deep learning subset of ML.
Can anyone recommend an online course for practically “learning AI”?
I’ve tried the hugging face course but got discouraged at the not quite working examples and colab books. There is also the Amazon and MS courses but I’d rather learn in a neutral way rather than a vendor-centric way.
Both Karpathy and fast.ai are good resources. Karpathy is "bottom up" (start from first principles and build on it) while fast.ai is "top down" (you start w/ working examples and gradually "peel off" to understand it).
Both approaches have merit, and it may come down to personal preferences.
I have looked through Prince's book (the free PDF) and think it is reasonably modern and pretty good. I'll probably pick up a copy at some point.
I have the new d2l book. At ~25 USD it is hard not to recommend it... the paperback print (color!) quality is good. I prefer reading print books so, for me, it is a good companion to the website, which is awesome.
I looked at the ToC and clicked into some of the chapters. It seems like this is more like a "Fundamentals of Machine Learning" book, as a lot of it is not really specifically about Deep Learning. Chapters that specific to Deep Learning seem to be 6-10, 12, 13 and 17-20. Nothing wrong with the contents, but I think the title is a bit misleading..
36 comments
[ 1.9 ms ] story [ 83.4 ms ] threadDoes anyone see a book icon? Or are we meant to flip through a slideshow embedded in the website?
Now that Amazon has competition they are not offering even $1 of discount. In the old days you'd get at least $20 off list price.
Price fixing at Amazon with books also included publishers: https://www.nytimes.com/2022/01/26/technology/amazon-price-f...
It is a new book, so my question is, is this title worth 90 bucks? Nope probably not. Maybe 9.99 for an ebook would be reasonable.
To paraphrase: Do you want piracy, because this is how you get piracy...
And it's available to read for free online, in its entirety, on the author's website.
It's hard to beat this proposition.
IMO, yes. I'll be ordering a copy. But "different strokes for different folks" and all that.
I also got Nathan Ida's book on Electromagnetic, hardcover, for ~40 USD direct from springer.
I'm about 10 feet from PRML right now, more than a decade after I got it.
see https://www.microsoft.com/en-us/research/publication/pattern... for the free PDF copy.
https://mml-book.github.io/book/mml-book.pdf
Besides this book are there any in the same league that are applicable to learn more about the diffusion and transformer model architectures?
https://web.stanford.edu/~jurafsky/slp3/
https://udlbook.github.io/udlbook/
[1] https://probml.github.io/pml-book/
I’ve tried the hugging face course but got discouraged at the not quite working examples and colab books. There is also the Amazon and MS courses but I’d rather learn in a neutral way rather than a vendor-centric way.
For reference, this is the course: https://karpathy.ai/zero-to-hero.html
I'm avoiding Azure or Google courses for the same reason. FastAI references its own (free) library but you can also do without it if you really try.
Both approaches have merit, and it may come down to personal preferences.
It has now been 8 years since I went through the Elements and Bishop Books in Uni. Now I want to read this over the Christmas break.
I'm interested in your opinion about them; both have pytorch code (notebooks).
___________________________
Understanding Deep Learning
by Simon J.D. Prince
Published by MIT Press Dec 5th 2023.
https://udlbook.github.io/udlbook/
https://www.amazon.com/Understanding-Deep-Learning-Simon-Pri...
___________________________
Dive into Deep Learning
https://d2l.ai/
https://www.amazon.com/Dive-into-Learning-Aston-Zhang/dp/100...
I have the new d2l book. At ~25 USD it is hard not to recommend it... the paperback print (color!) quality is good. I prefer reading print books so, for me, it is a good companion to the website, which is awesome.
Series: https://www.statlearning.com