- This event has passed.
Tor LATTIMORE (DeepMind London) – ” Improved Regret for Zeroth-Order Adversarial Bandit Convex Optimisation”
September 28, 2:00 pm - 3:15 pm
The Statistical Seminar: Every Monday at 2:00 pm.
Time: 2:00 pm – 3:15 pm
Date: 28th of september 2020
Tor LATTIMORE (DeepMind London) – “Improved Regret for Zeroth-Order Adversarial Bandit Convex Optimisation“
Abstract: I will explain recent advances in adversarial bandit convex optimisation, improving the best known minimax regret from O(d^(9.5) sqrt(n) log(n)^(7.5)) to O(d^(2.5) sqrt(n) log(n)) where d is the dimension and n is the number of interactions. The new analysis combines minimax duality with the information-theoretic Bayesian regret analysis by Russo and Van Roy as well as basic tools from asymptotic convex geometry.
A preprint of this work is available at https://arxiv.org/pdf/2006.00475.pdf.
Cristina BUTUCEA (CREST), Alexandre TSYBAKOV (CREST), Karim LOUNICI (CMAP) , Zoltan SZABO (CMAP)