What is the best way to start machine learning? I know Python

Have you learn data structure, oop, recursion, and basic AI?

The best way is google and a looot of time

First : get one billion different images of the same thing!
Second : Enjoy your retirement! 😃

@volttide those are all nice-to-haves, not really relevant from a pure machine learning viewpoint.

@indraasura it depends on what level you want to learn ML at. Simple models and stuff, you could just look up what each algorithm does intuitively (eg. PCA does dimensionality reduction by finding new data axes and projecting your dataset along those, or linear regression fitting lines to the dataset, SVMs, random forest etc.). Then just use a library like scikit-learn and build you models.

The problem with the above approach is that you'll only get so far, you'll hit a skill ceiling pretty fast. But it is sufficient for basic stuff.

If your goal is to learn it in depth, check out Andrew Ng's Stanford course on ML on YouTube. It's pretty heavy though. Ng's Coursera course is much easier and is a decent introduction (eg. Here you'll learn that PCA is done via a linear algebra operation called the singular value decomposition and how it works).

To start learning how to get the machine to learn you must first learn from the machine on how to get the machine to learn.

Does that make sense?