Model and Policy Robustness in Online Optimization (ROPT), Spring-Summer 2022
Objective
The objective is to learn about recent research that examines various robustness issues at the intersection of online optimization and machine learning.
Summary
Session 1: Basics of Multi-Armed Bandits (Scribe: Alp Sungu)
Session 2: Information-Theoretic Analysis of MAB Policies (Scribe: Naireet Ghosh)
Session 3: Robustness of Thompson Sampling (Scribe: Tong Wang)
Session 4: Transfer Learning I (Scribe: Alireza Amanihamedani)
Session 5: Transfer Learning II (Scribe: Alireza Amanihamedani)
Session 6: Fairness in Exploration (Led by Omer and Alireza) May 3
Session 7: Blackwell Approachability Theorem and Applications I May 10
Session 8: Application of MABs to vaccine allocation (Led by Tong and Naireet) May 24
Session 9: Blackwell Approachability Theorem and Applications II (Scribe Yuhang) June 7
Session 10: Adaptivity and Confounding in MABs June 14
Session 11: Short research presentations June 21
List of papers (upcoming sessions):
Student-lead sessions:
MABs with Fairness (S6)
Thompson Sampling and Vaccine Allocation (S8)
Additional Bandit Applications to Healthcare [1][2] (S8)
Deconfounding in MABs (S10)
Ali's sessions:
Blackwell approachability (S7-9)
Additional readings:
Online matching with ML advice
ML predictions for online optimization
Participants:
Alp Sungu, Omer Saritac, Tong Wang, Naireet Ghosh, Baizhi Song, Yuhang Du, Alireza Amanihamedani, Allen Li, Jean Pauphilet, Ali Aouad.
Time and Location:
5pm-7pm, LT3 unless mentioned otherwise.