@(Cabinet)[how_to_study_algo|published]

180 Coding Experiments in 180 Days

start: 2015-02-18

Write Code for Machine Learning Every Day

I create a project called 180 coding experiments in 180 days in my GitHub.

My confession

Recently, I really feel that my poor programming skills seriously limit my machine learning experiments and understandings. For example, I've been struggling with the code of PyLearn2 and Spearmint for the last several months, but without too much improvement.

Then I started from easier ones such as the code in Machine Learning: An Algorithmic Perspective, which makes me feel good. By following through the code flows, it is easier to understand some deep concepts in machine learning. I now believe that writing and reading code is essential to grasping the ideas of machine learning algorithms.

Code as habit

Inspired by the post by John Resig, I decide to set several rules:

1. I must write code every day. I can read books, or write blog posts, or other things but it must be in addition to the code that I write.
2. It must be useful code. No tweaking indentation, no code re-formatting, and if at all possible no refactoring. (All these things are permitted, but not as the exclusive work of the day.)
3. The code must be in GitHub, preferably in the form of IPython Notebook.

Inspired by the 180 websites in 180 days project by Jennifer Dewalt, I create a project called 180 coding experiments in 180 days in my GitHub.

There are some useful tools, such as HabitRPG, Chains.cc, WakaTime, suggested in a post of Hacker News.