BLISS Seminar

About

The BLISS seminar (formerly NCD seminar) is the area seminar of the Berkeley Laboratory for Information and System Sciences. Talks at the seminar cover topics including but not limited to information and coding theory, signal processing, optimization, statistics, and control. The list of talks for the current semester can be found below, and past seminars from 2016 onwards are listed here. For an archive of all talks from 1996-2015, visit the old webpage.

A calendar of all the talks is maintained here. Feel free to add it to your own.

Fall 2022
Location: Virtual (subscribe to our mailing list for details)
Regular seminar time: Mondays 3 PM - 4 PM PT
Regular seminar venue: Hughes Conference Room, 400 Cory Hall, Berkeley (map)

To subscribe to our mailing list, click here.

To give a talk at the seminar, contact Nived Rajaraman or Tom Courtade.

Recordings of the talks from Fall 2022 are available here.
Recordings from Spring 2022 are available here.

Fall 2022 Talks

Dates marked in bold indicate that talks are at non-regular dates / times.

Friday, August 26, 3-4 PM Weijie Su (U Penn) When Will You Become the Best Reviewer of Your Own Papers? A Mechanism-Design-Based Approach to Statistical Estimation details
August 29 Haipeng Luo (USC) Near-Optimal No-Regret Learning for General Convex Games  —  The Role of Positive Regret details
Oct 10 Yuxin Chen (U Penn) Towards Optimal Sample Complexities in Offline Reinforcement Learning and Markov Games details
Oct 17 Wen Sun (Cornell) Hybrid RL: Using both offline and online data can make RL efficient details
Oct 24 Chen-Yu Wei (Simons) Optimal Dynamic Regret for Bandits without Prior Knowledge details
Oct 31 Wenlong Mou (UC Berkeley) Rethinking semi-parametric efficiency for off-policy estimation: a non-asymptotic perspective details
Tuesday, Nov 1, 2-3 PM Kevin Jamieson (University of Washington) Towards Instance-Optimal Algorithms for Reinforcement Learning details
Nov 14 Raaz Dwivedi (Harvard and MIT) Two vignettes on efficient procedures for personalized decision making details
Nov 28, 2-3 PM Gautam Kamath and Mahbod Majid (University of Waterloo) Efficient Mean Estimation with Pure Differential Privacy via a Sum-Of-Squares Exponential Mechanism details