I have a question and I am not sure if anyone have an answer for it but I appreciate any insights, thoughts, or experiences you might have. I am a CS major and I have passed
Calculus I, II , III
Probability and statistic for engineering
Applied combinatorics
Linear algebra
Numerical linear algebra
Differential equations
Algorithm analysis and design
Machine learning
I may not proficient in all these subjects(since I passed some of them 10 years ago) but I know where to look for the answers when I encounter a simple problem that is related to one of these subjects.
I have done proof problem for important theorem in those classes. But I am not convenient in reading CS paper that has a lot of mathematical proof and of course I don't know how to write such papers. As an example when I read this paper I could understand up to the end of section 2. In section 3 when it starts having assumptions, theorem and proof, I get frustrated and I cannot understand the paper well. Probably If I spend 4 days, I will get it. I have encountered a lot of such papers are in CS or IE( for example this paper). I expected the courses I passed(listed above), help me to understand them in a reasonable amount time.
I want to start doing research in machine learning but I don't know I am ready to even choose a topic. My question is I struggled in reading such papers because :
1- I have not read enough related paper. It is natural to spend great amount of time for a non-math major. So I should spend days for the first few papers and I can improve later.
2- The courses I listed are not enough and I need to study more for having the skill to read and write such papers. If that is the case do you have any suggestions what class should I take to help me to improve?
3- I should find an advisor that published such papers, taking her/his classes and If I passed the class, hopefully I will be ready to understand them.
What is the best strategy to improve? Thanks in advance.
Have you gone through any machine learning textbooks?
Books like K.P. Murphy's Machine Learning: A Probabilistic Perspective have reasonably up to date coverage of a lot of topics, and are a good place to start with getting up to the basics of the field (for example, the paper you link on spectral clustering has been textbook / tutorial material for a while).
Larry Wasserman's All of statistics is a useful supplement.
After that, reading papers is somewhat about maturity and takes practice -- you can get the ideas of the papers faster if you've seen more of them.