Mittagsseminar (in cooperation with J. Lengler, A. Steger, and D. Steurer)

Date and Time: Friday, May 03, 2019, 12:15 pm

Duration: 30 minutes

Location: CAB G 11

Speaker: Thodoris Lykouris (Cornell University)

Small-loss bounds for online learning with partial information

I will discuss the problem of online learning with graph-based feedback in the adversarial (non-stochastic) setting. At each round, a decision maker selects an action from a finite set of alternatives and receives feedback based on a combinatorial graph-based feedback model introduced by Mannor and Shamir. This encapsulates as special cases important partial information paradigms such as bandits, online routing, and contextual bandits. An important challenge in such partial-feedback settings is that, in order to keep up with the non-stochasticity in the environment, the learner needs to explore often all actions, including suboptimal ones. I will tackle this hurdle by providing a general black-box reduction to attain effective regret guarantees that avoid this over-exploration. Joint work with Karthik Sridharan and Eva Tardos

Upcoming talks     |     All previous talks     |     Talks by speaker     |     Upcoming talks in iCal format (beta version!)

Previous talks by year:   2025  2024  2023  2022  2021  2020  2019  2018  2017  2016  2015  2014  2013  2012  2011  2010  2009  2008  2007  2006  2005  2004  2003  2002  2001  2000  1999  1998  1997  1996  

Information for students and suggested topics for student talks

Automatic MiSe System Software Version 1.4803M   |   admin login