So What Is Online Convex Optimization Anyway?

It's fascinating and a lot of fun!

The content of this post is pretty much the word-for-word explanation of the subject matter in chapter 1 of Introduction to Online Convex Optimization by the author Elad Hazan himself. This is no accident — as briefly mentioned in my “zeroth” post, this book is amazingly well-written and captivates the reader’s attention immediately.

I could have simply screenshotted or typed up the relevant excerpts and presented them here as-is, but in my opinion that would have been lazy and almost disrespectful to the author. Instead, I have taken myself to task and expended a bit of my own effort by taking some hand-written notes and highlighting points that I found particularly impactful, which should make this post transformative enough to warrant its existence. 🔥😎🎉

1. The Online Convex Optimization Setting

2. Restrictions Required by the OCO Framework

3. Why These Restrictions are Necessary

---

And voilà! The above is of course but the tip of the iceberg and does not, on its own, provide the concrete implementation details of the OCO protocol, but worry not — I’ll delve into all that in a future post. I’m trying to be bite-sized and not overly mathematical here, because for one I’ve been told I tend to ramble a lot and people simply tune out after a while. Well, my brain is capable of incorporating critical but constructive feedback, so I hope this post is a good demonstration of that.

🔥😇🎉

Yours truly,

A Fan of OCO