[joker3]

This is nice to have. The classic reference on online learning is "Prediction, Learning and Games" but it's not ML focused and it predates pretty much everything on online convex optimization. I'm looking forward to seeing a more modern introduction.

[bhl]

What's the difference between ML and online convex optimization? I remember presenting on the Multiplicative Weights algorithm, and someone told me not to introduce it from a learning theory perspective and instead from more of a basic theoretical CS perspective -- that part was confusing, because they seemed to me to be the same.

[joker3]

Take a look at https://ocobook.cs.princeton.edu/OCObook.pdf, and in particular section 1.2. The online shortest path problem is a good example of an online optimization problem that's not related to machine learning.