ML-Tips.md

1. ML takes time.
2. Math : Make a math toolkit of definitions and rules and learn what to use when for mathematical reasoning and derivations for algorithms by practice. Now instead of looking at a formula (say loss func for Decision Tree is information gain) as a whole bunch of symbols and numbers, think about what you want to achieve in plain text and then this final formula is the result of a series of steps equivalent to pseudo code we write for code
3. Code : writing code is not actual coding, debugging a code is. It might take 1 hour to write 5 lines of ML code and another 3 hours to debug it, and this is expected and normal.
4. To start working on an existing large code base: do not search all files in all the folders randomly, begin with train.py and eval.py and follow the function loops and flow from there.

References : https://www.youtube.com/watch?v=sJBO7rMR8ks