0
$\begingroup$

There are multiple books on deep learning currently available. I'm primarily interested in the theory and algorithms and less interested in the "practical guide" books really just tell me how to use a specific deep learning library (like PyTorch or TensorFlow) and black-box the mathematical derivations. I've seen various books recommended to me:

  • Deep Learning by Aaron C. Courville, Ian Goodfellow, and Yoshua Bengio
  • Online book at http://neuralnetworksanddeeplearning.com by Michael Nielsen
  • Neural Networks and Learning Machines, 3rd edition, by Simon Haykin
  • Neural Networks: A Systematic Introduction by Raúl Rojas

I've read through the Courville-Goodfellow-Bengio book, and I though it covered a wide variety of topics. I'm wondering if it's still worth reading the other books. Could someone who has read these books do a compare-and-contrast on them? Particularly, I'm wondering if its still worth reading the other books given what I have already read, or if reading them would just go over content that I've already seen in the first book and be redundant.

$\endgroup$
  • $\begingroup$ I think putting into practice what you have learned thus far, will inform your decision about what you need going forward. $\endgroup$ – grldsndrs Jul 5 at 2:53
  • $\begingroup$ I've learned how to use PyTorch, and I've participated in some image classification competitions on Kaggle. I didn't find "putting into practice" my knowledge particularly enlightening largely I'm just using the same standard data science practices that are taught in every introductory, undergraduate data science courses. It also doesn't particularly feel like I'm doing state-of-the-art work because the interesting mathematics is hidden behind the code, and all the work is experimental. This is why I'm asking for more theoretical texts. $\endgroup$ – Anon Jul 5 at 3:19
  • $\begingroup$ Have you tried implementing the "standard data science practices" from scratch?: not relying on the prefabricated functions prevalent in DS and ML packages. I suspect you will find quickly where you could be more informed. At any rate, I found that 'Pattern Classification' by Duda, Hart, and Stork covered the maths and underlying reasoning relatively well. $\endgroup$ – grldsndrs Jul 5 at 3:37
  • $\begingroup$ Also, the main question I'm asking here is a compare-and-contrast between these books. I'm sure that would be a valuable question to answer for both me and future SE users. $\endgroup$ – Anon Jul 5 at 3:39
  • $\begingroup$ Yes, I have implemented "standard data science practices" from scratch. They aren't that deep or even mathematically interesting. It's entirely experimental. I appreciate the additional book recommendation, but that's not what this question is really asking about. Based on the ToC of "Pattern Classification" by Duda, Hart, and Stork, I don't think that's what I am looking for either. I'm asking about specifically deep learning books, not books on general machine learning that also happens to talk about deep learning. I've already read both the Bishop and Murphy textbooks for those topics. $\endgroup$ – Anon Jul 5 at 3:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.