1
$\begingroup$

I am trying to read the following list of books on statistical learning. I have a BSCS and about 4 yrs of experience working in image processing and parallel programming. I won’t be an expert in the field by any means, however my aim is:

  • not to be a script kiddie, using tools and algorithms without understanding the hows and whys.
  • be able to read and digest the latest research in statistical learning, specially w.r.t computer vision.

Prerequisites I have studied in preparation:

Books to read:

  • An Introduction to Statistical Learning with Applications in R by Robert Tibshirani et al.
  • The Elements of Statistical Learning: Data Mining, Inference, and Prediction by Robert Tibshirani et al.
  • Understanding Machine Learning From Theory To Algorithms by Shai Shalev-Shwartz et al.
  • Pattern Recognition and Machine Learning by Christopher M. Bishops
  • Information Theory, Inference, and Learning Algorithms by David J.C. MacKay
  • Deep Learning by Ian Goodfellow et al.
  • Convex Optimization by Stephen Boyd et al.

I am looking for a reading strategy. I'd specially appreciate input from users who’ve read the majority of the books.

$\endgroup$
2
$\begingroup$

I've read all but the last one. If you are serious about computer vision you won't get much out of this reading list. However, if you are set on these books I would start with Goodfellow's Deep Learning which is fairly self-contained and the only book that covers computer vision enough to make it rewarding for that purpose. You can branch out from here and read additional material referenced in the book where you lack background.

If you are less concerned with computer vision you can start with the other books, they are often read in the following order:

  1. An Introduction to Statistical Learning with Applications in R
  2. The Elements of Statistical Learning
  3. Bishop's Pattern Recognition and Machine Learning

If you finished with these you could tackle McKay which is quite dry and boring. Shai Shalev-Shwartz et al. is a nice book that covers a lot of the material in the other books but from the perspective of Empirical Risk Minimization and covers Vapnik–Chervonenkis theory. It is more theoretical and is more structured in the classical theorem-proof-corollary scheme. The perspective taken is less practical but adds to the other books.

I don't know the book on Convex Optimization but you can probably read it in any order. If you want to read all of the books and don't care how long it takes, you can probably read that first as it sounds more foundational.

| improve this answer | |
$\endgroup$
  • $\begingroup$ This is exactly what I was looking for. Even though end goal is CV, I want to build a solid foundation in ML. What would you recommend for someone serious about CV? $\endgroup$ – zindarod Oct 11 '18 at 18:33
  • $\begingroup$ Difficult question... probably not the right question for this forum. Computer vision is a big field and people nowadays seem (for understandable reasons) mostly interested in recognition using deep learning. If you want to get an overview of the depth and breadth of the field, there are a couple books out there but I can't comment on which one is best. One I'm aware of is from Szeliski which has a free online version at szeliski.org/Book $\endgroup$ – oW_ Oct 11 '18 at 21:16
1
$\begingroup$

I would start directly with a Stanford course of image recognition with neural networks (http://cs231n.stanford.edu/syllabus.html). They have slides that are fast to read and easy to understand. They have an overview of the methods used in image recognition and also have a detailed description of the latest research (updated every year). They also cover deep learning and gradient descent methods. I would also advise you to get a good book on statistics that explains among others CLT, parameters vs overfitting dilemma and CV. And take part in a Kaggle competition. This will allow you to understand the practical limitations of current methods, get to know current state of the art methods in image recognition and see how people solve real world problems. Here is an image recognition kaggle competition running now: https://www.kaggle.com/c/airbus-ship-detection . After the Stanford course slides and a Kaggle competition you will probably be able to understand almost all of scientific papers on image recognition and computer vision.

| improve this answer | |
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.