@PatrickKidger Good advice! I like your list, but there are other important topics IMHO (variational inference, SMC, GPs (despite your aversion), etc). Maybe make this into an editable doc on github, so ppl can add links?

@PatrickKidger Yeah I agree. And also especially that you often don't just say "learn about X" but rather "implement a small X from scratch" - I think this is the crucial part that makes all the difference. Watching a YouTube lecture about X is not enough to actually know X !

@PatrickKidger That is quite a list! On the (theoretical) statistics side, I would probably add some statistical learning theory: Boosting, no free lunch and VC dimensions are important topics from the early days of ML. Also, if you're into RL, bandits and MDPs would probably help

@PatrickKidger I would put a lot of weight on finding the right advisor, lab, and university. This is arguably harder to get right and even harder to correct. Seems like a lot of luck is involved.

@PatrickKidger How would you recommend a masters student (thesis based) approach this?

@PatrickKidger i want to read more about why you don't like gps !

@PatrickKidger I think this is a fantastic list. I think I know about 60% of the stuff in here and it's a great launching place for the last 40%.

@PatrickKidger thanks! incredibly useful list. JKS is probably the only thing that actually matters.

@PatrickKidger @readwise save thread

@PatrickKidger A great list. However, which are the best courses or books etc, to learn from? I think that recommendations in this direction would make the list even better.

@PatrickKidger I think the idea of a cumulative over time "Just Know Stuff" list is such a cool idea if different people within the field would share pieces of core knowledge they think are useful