Deep Reinforcement Learning

Program

Theory of Reinforcement Learning

Date

Monday, Sept. 28 – Friday, Oct. 2, 2020

Back to calendar

All talks listed in Pacific Time (PDT).

8:50 – 9 a.m.

Opening Remarks

9 – 9:30 a.m.

MOPO: Model-Based Offline Policy Optimization

Tengyu Ma (Stanford)

9:30 – 10 a.m.

Learning from the Past Without Great Exploration

Emma Brunskill (Stanford)

10 – 10:30 a.m.

Offline Deep Reinforcement Learning Algorithms

Sergey Levine (UC Berkeley)

10:30 – 11 a.m.

Break

11 – 11:30 a.m.

Attacking the Off-Policy Problem With Duality

Ofir Nachum (Google Research)

11:30 a.m. – 12 p.m.

Discussion: Offline Reinforcement Learning

Pablo Castro (Google)

12 – 1 p.m.

Gather.town

9 – 9:30 a.m.

Behaviour Suite for Reinforcement Learning

Ian Osband (DeepMind)

9:30 – 10 a.m.

Learning Exploration Strategies with Meta-Reinforcement Learning

Chelsea Finn (Stanford University)

10 – 10:30 a.m.

Temporally-Extended ε-Greedy Exploration

Will Dabney (DeepMind)

10:30 – 11 a.m.

Break

11 – 11:30 a.m.

PC-PG: Policy Cover Directed Exploration for Provable Policy Gradient Learning

Alekh Agarwal (Microsoft Research)

11:30 a.m. – 12 p.m.

Discussion: Exploration

Joel Lehman (Uber)

12 – 1 p.m.

Gather.town

9 – 9:30 a.m.

Munchausen Reinforcement Learning

Matthieu Geist (Google)

9:30 – 10 a.m.

Adaptive Approximate Policy Iteration

Nevena Lazic (DeepMind)

10 – 10:30 a.m.

Stabilizing Q-learning with Weighted Bellman Losses

Pieter Abbeel (UC Berkeley)

10:30 – 11 a.m.

Break

11 – 11:30 a.m.

Generalizing the Projected Bellman Error Objective for Nonlinear Value Estimation

Martha White (University of Alberta)

11:30 a.m. – 12 p.m.

Discussion: Optimization

Gergely Neu (UPF)

12 – 1 p.m.

Gather.town

9 – 9:30 a.m.

Language as a Scaffold for Reinforcement Learning

Jacob Andreas (MIT)

9:30 – 10 a.m.

Invariant Prediction for Generalization in Reinforcement Learning

Clare Lyle (University of Oxford)

10 – 10:30 a.m.

Exploiting Latent Structure and Bisimulation Metrics for Better Generalization

Amy Zhang (McGill University, Mila Institute, Facebook AI Research)

10:30 – 11 a.m.

Break

11 – 11:30 a.m.

Fast Reinforcement Learning With Generalized Policy Updates

Doina Precup (McGill Univeristy & MILA / DeepMind)

11:30 a.m. – 12 p.m.

Policy Gradients Methods, Neural Policy Classes, and Distribution Shift

Sham Kakade (University of Washington)

12 – 1 p.m.

Gather.town

9 – 9:30 a.m.

Is Safe Learning the Future of RL?

Scott Niekum (UT Austin)

9:30 – 10 a.m.

Rigorous Uncertainty Quantification for Off-policy Evaluation in Reinforcement Learning: a Variational Approach

Qiang Liu (UT Austin)

10 – 10:30 a.m.

CoinDICE: Off-Policy Confidence Interval Estimation via Dual Lens

Bo Dai (Google Brain)

10:30 – 11 a.m.

Break

11 – 11:30 a.m.

Mixed Autonomy Traffic: A Reinforcement Learning Perspective

Cathy Wu (MIT)

11:30 a.m. – 12 p.m.

Deep Robust Reinforcement Learning and Regularization

Shie Mannor (Technion)

12 – 1 p.m.

Gather.town