Studying machine learning

Question

I'm currently studying machine learning using Bishop's book "Pattern Recognition and Machine Learning". The main disadvantage of this book (for me) is a lack of practical applications. Also it seems like some chapters somehow replicate each other, following the same structure: cons of MLE approach, pros of Bayesian approach, Laplace approximation and some (very superficial) information. Thus, it could be reasonable to read either only the last few sections of each chapter or to find out the main idea by the first few sections and skip the rest ones since they are straightforward.

On the other hand, I found books like "ML for (programming language): ..." focus on implementation of algorithms rather than explanation of choosing an appropriate algorithm for a given problem. Therefore, if one cuts out programming specific topics the book becomes very poor.

I want to understand theoretical foundations of basic ML algorithms and learn how to apply these algorithms in practice. While Bishop's book covers the first goal fairly well, information for the second one is not enough. Since I just briefly flipped through other books I could get a wrong impression about them. So can anyone please suggest a book(s) or some other resource that fits my needs and/or complement Bishop's book?

score 3 · Answer 1 · answered Jun 19 '15 at 14:11

3

Feel free to check out Tom Mitchell's Machine Learning. I think it covers the foundations pretty well, and once you have exposure to the different approaches, your intuition grows as to which makes more sense for a given ML task. Plus, later chapters build on earlier chapters, making this building easier.

answered Jun 19 '15 at 14:11

Ken

3,751

2

I have found a cheat-sheet which is something like what I am looking for. A kind of such diagram with detailed mathematical explanation and some practical examples will perfectly fit my needs. – vladkkkkk Jun 22 '15 at 22:47
@nmerci Wonderful! – Ken Jun 23 '15 at 01:43

score 3 · Answer 2 · answered Jun 30 '17 at 01:54

To gain exposure to applications of machine learning, I highly recommend a library of algorithms implemented in Matlab to accompany Bishop's PRML book: https://github.com/PRML/PRMLT

For additional textbooks, Kevin Murphy's "Machine Learning: A Probabilistic Perspective": https://www.cs.ubc.ca/~murphyk/MLbook/ is an excellent additional resource for both theory and practice of machine learning.

Studying machine learning

2 Answers2

Linked