Simple, Representative Concrete Examples

When an algorithm or process feels magical, that's typically an indication you don't really understand what's happening under the hood.

One of the most frustrating things about existing ML learning resources is a lack of simple yet representative concrete examples.

Given any ML topic, existing resources will typically just present some algorithm or process WITHOUT a concrete example –

and if there IS a concrete example, then it’s typically either

• so simple that you don't know WTF it's trying to demonstrate, or
• so complicated that you don't know WTF is going on. You just end up running some code and thinking "wow, that was a cool magical incantation."

A concrete example should ideally be

• representative, i.e., it encapsulates the important contours of the problem being solved, and
• simple enough that you can track important steps and quantities at a fully granular level.

You shouldn’t have to imagine or infer anything – the info should be right there in front of you.

When an algorithm or process feels magical, that’s typically an indication you don’t really understand what’s happening under the hood.

There’s something you’re supposed to be imagining, but you don’t have it in your head, so the outcome just feels like magic.