Successor Uncertainties: Exploration And Uncertainty In Temporal Difference Learning

Abstract

Posterior sampling for reinforcement learning (PSRL) is an effective method for balancing exploration and exploitation in reinforcement learning. Randomised value functions (RVF) can be viewed as a promising approach to scaling PSRL. However, we show that most contemporary algorithms combining RVF with neural network function approximation do not possess the properties which make PSRL effective, and provably fail in sparse reward problems. Moreover, we find that propagation of uncertainty, a property of PSRL previously thought important for exploration, does not preclude this failure. We use these insights to design Successor Uncertainties (SU), a cheap and easy to implement RVF algorithm that retains key properties of PSRL. SU is highly effective on hard tabular exploration benchmarks. Furthermore, on the Atari 2600 domain, it surpasses human performance on 38 of 49 games tested (achieving a median human normalised score of 2.09), and outperforms its closest RVF competitor, Bootstrapp

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyjanz2018successor

Related papers

• Why Is Posterior Sampling Better Than Optimism For Reinforcement Learning? (2016)0.00
• Accelerating Residual Reinforcement Learning With Uncertainty Estimation (2025)0.00
• Posterior Sampling With Delayed Feedback For Reinforcement Learning With Linear Function Approximation (2023)0.00
• Count-based Exploration With The Successor Representation (2018)13.17
• Uncertainty Quantification And Exploration For Reinforcement Learning (2019)6.77
• You Can't Count On Luck: Why Decision Transformers And Rvs Fail In Stochastic Environments (2022)0.00
• Posterior Sampling For Large Scale Reinforcement Learning (2017)0.00
• Model-based Reinforcement Learning For Continuous Control With Posterior Sampling (2020)0.00