How do I Learn Machine Learning?

Here’s how I did it:

• take every single online course you can involving data science. That includes basic ML, linear algebra (probably the most important), statistics, visualization, deep learning etc.

• read constantly. read blogs, classic texts, journal articles. I kept books by my bed and would read every night until I passed out.

• Not only read this stuff, but master it. I’d get a new paper or book chapter I liked and I’d go to a coffee shop and rewrite the entire thing by hand w pencil and paper. I’d go through every derivation and memorize it. Any time a foreign concept was referenced, I’d research it and do the same thing with that concept - rewrite everything I read by hand, including my own thoughts, master it, then go back to the original paper

• keep learning and relearning everything until it was second nature. I probably learned and relearned something as simple as linear regression dozens of times, each time understanding it from a new perspective.

• do the same thing with programming. I’d consider myself a nearly complete expert in Python. You need to be a quick and effective programmer to be of any use. I know at least one very talented and knowledgeable data scientist whose mediocre programming skills make him almost useless.

• Read about new tools and packages for your language of choice every day. Make a conscious effort to integrate them into my toolbox if it would help me reduce an expression of a procedure by even one line of code. It all adds up, your power multiplies.

• Do as many kaggle competitions as possible. Work on as many datasets as you can, and understand how all the classic ML algorothms would interact with the structure of that data. Each dataset had its own idiosyncracies; get familiar with a wide variety of data.

• Get a job somewhere that works with ML. At first, they might not let you touch the ML. Make ML stuff that would aid the business in your spare time to show them that you know your stuff and can implement it. They’ll probably let you start working on ML.