Convergence and Sample Complexity of Gradient Methods for the Model-Free Linear Quadratic Regulator

Channel:
United States Simons Institute for the Theory of Computing
Subscribers:
68,700
Published on ● Video Link: https://www.youtube.com/watch?v=iiE4ijPf27Y


Duration: 32:36
880 views
14

Mihailo Jovanovic (USC)
https://simons.berkeley.edu/talks/tbd-255
Reinforcement Learning from Batch Data and Simulation



Other Videos By Simons Institute for the Theory of Computing


2020-12-09Price of Active Security in Multiparty Computation
2020-12-08Game-Theoretically Secure Protocols Inspired by Blockchains
2020-12-08Lightning Network Economics: Cost-Minimal Channels and their Implications for Network Structure
2020-12-08Local Proofs Approaching the Witness Length
2020-12-08Lower Bounds for Off-Chain Protocols: Exploring the Limits of Plasma c
2020-12-08Central Bank Digital Currency: When Price and Bank Stability Collide
2020-12-08Blockchain as Regulatory Technology: from Code is law to Law as Code
2020-12-07Subquadratic SNARGs in the Random Oracle Model
2020-12-05Panel Discussion
2020-12-05Offline Reinforcement Learning and Model-Based Optimization
2020-12-04Convergence and Sample Complexity of Gradient Methods for the Model-Free Linear Quadratic Regulator
2020-12-04Policy Evaluation under Interference
2020-12-04Stable Reinforcement Learning with Unbounded State Space
2020-12-04Multiagent Reinforcement Learning: Rollout and Policy Iteration
2020-12-04Sample Efficient Reinforcement Learning via Low-Rank Matrix Estimation
2020-12-04Nearly Minimax Optimal Reward-Free Reinforcement Learning
2020-12-03Statistical Efficiency in Offline Reinforcement Learning
2020-12-03Batch Policy Learning in Average Reward Markov Decision Processes
2020-12-03Panel Discussion
2020-12-02The Mean-Squared Error of Double Q-Learning
2020-12-02Q-learning with Uniformly Bounded Variance


Tags:
Simons Institute
theoretical computer science
UC Berkeley
Computer Science
Theory of Computation
Theory of Computing
Mihailo Jovanovic
Reinforcement Learning from Batch Data and Simulation