Corpus ID: 67787873

Statistics and Samples in Distributional Reinforcement Learning

@inproceedings{Rowland2019StatisticsAS,
  title={Statistics and Samples in Distributional Reinforcement Learning},
  author={Mark Rowland and Robert Dadashi and S. Kumar and R. Munos and Marc G. Bellemare and Will Dabney},
  booktitle={ICML},
  year={2019}
}
We present a unifying framework for designing and analysing distributional reinforcement learning (DRL) algorithms in terms of recursively estimating statistics of the return distribution. Our key insight is that DRL algorithms can be decomposed as the combination of some statistical estimator and a method for imputing a return distribution consistent with that set of statistics. With this new understanding, we are able to provide improved analyses of existing DRL algorithms as well as… Expand
16 Citations
Distributional Reinforcement Learning via Moment Matching
  • Highly Influenced
  • PDF
Distributional Reinforcement Learning with Maximum Mean Discrepancy
  • Highly Influenced
Bayesian Distributional Policy Gradients
  • Luchen Li, A. Faisal
  • Computer Science
  • ArXiv
  • 2021
  • PDF
Learning to Score Behaviors for Guided Policy Optimization
  • 8
  • PDF
A Local Temporal Difference Code for Distributional Reinforcement Learning
  • 3
  • Highly Influenced
  • PDF
The Value-Improvement Path: Towards Better Representations for Reinforcement Learning
  • 6
  • PDF
...
1
2
...

References

SHOWING 1-10 OF 20 REFERENCES
Distributional Reinforcement Learning with Quantile Regression
  • 164
  • PDF
Nonparametric Return Distribution Approximation for Reinforcement Learning
  • 61
  • PDF
Distributional reinforcement learning with linear function approximation
  • 12
  • PDF
An Analysis of Categorical Distributional Reinforcement Learning
  • 36
  • PDF
Parametric Return Density Estimation for Reinforcement Learning
  • 56
  • PDF
A Comparative Analysis of Expected and Distributional Reinforcement Learning
  • 32
  • PDF
A Distributional Perspective on Reinforcement Learning
  • 504
  • PDF
Nonlinear Distributional Gradient Temporal-Difference Learning
  • 8
  • Highly Influential
  • PDF
...
1
2
...