If you think logistic regression is machine learning, you are right.

If you think logistic regression is statistics, you are right.

If you think it can only be either or, you are wrong.

What would a minimal, SOTA supervised ML course look like?

- LLMs
- xgboost
- CNNs
- CV
- metrics
- random search (tuning)
- one-hot encoding
- Linear/logistic regr.

What else?

You shouldn't use over-and undersampling to "rebalance" your classes for machine learning.

Instead?

Do nothing.

Then you already have an edge over the oversamplers.

Took me a while, but I now have good systems that free up my brain for actual thinking and keep my short-term memory free of trivial stuff:

- Getting Things Done for ToDos
- Second Brain for notes
- Google Calendar for events

What systems help you manage your life?

Self-supervised learning is half supervised, half unsupervised.

Supervised because it actually works with labels, it's just that they are automatically created.

It's unsupervised because it leverages data to learn representations without the need for external annotations

I have a Bachelor's, a Master's, a PhD, and written 4 books.

But only very recently did I actually learn how to systematically take notes and manage references. 🤷‍♂️

Looks like I'm just getting started, my friends!