Francesco Orabona (@bremen79)

Francesco Orabona @bremen79

Associate professor at @BU_ece. Formerly at @sbucompsc, @YahooResearch, and @TTIC_Connect. I hate hyperparameters in ML algorithms and pineapple on pizza.

Tweets by Francesco Orabona

You cannot take the logarithm of the Lipschitz constant of a function! A 🧵 about a super common mistake in ML papers 1/10

Shared by Francesco Orabona at 5/10/2022 36-257

https://twitter.com/bremen79/status/1334173744038879235

Wow, it happened!!! "Optimization for Deep Learning" is not a subtopic in ICML 2022! Thanks @icmlconf!

Shared by Francesco Orabona at 1/19/2022 0-54

https://arxiv.org/abs/2007.14294

My student @xyli94 noticed it in her paper at Workshop on Beyond First Order Methods in ML Systems at ICML 2020: https://t.co/EE5aIMmaCK She al ...

Shared by Francesco Orabona at 1/19/2022 8-26

Anyone has this pdf? Kearns and Valiant. Learning boolean formulae or finite automata is as hard as factoring. Technical Report TR-14-88, Harva ...

Shared by Francesco Orabona at 12/6/2021 0-8

https://parameterfree.com/2021/12/02/saddle-point-problems-and-oco/

New blog post: Saddle-point problems and Online Convex Optimization What is a saddle-point? Does a saddle-point always exist? Can we use online ...

Shared by Francesco Orabona at 12/2/2021 32-179

Me: Studying some details on the book of Cesa-Bianchi&Lugosi My wife: "Are you reading the Bible again?"

Shared by Francesco Orabona at 12/2/2021 0-74

Let's plainly admit it: Some reviews are just pure non-sense Now, there is an interesting side-effect of the @iclr_conf system: some authors ar ...

Shared by Francesco Orabona at 11/27/2021 2-53

https://www.tensorflow.org/addons

After 4 years,the COCOB algorithm to train deep networks without learning rates is part of TensorFlow! *All* the merit goes to @rehan_guha that ...

Shared by Francesco Orabona at 7/13/2021 34-136