• Mathematics, Computer Science
  • Published in COLT 2019

An Information-Theoretic Approach to Minimax Regret in Partial Monitoring

@inproceedings{Lattimore2019AnIA,
  title={An Information-Theoretic Approach to Minimax Regret in Partial Monitoring},
  author={Tor Lattimore and Cs. Szepesvari},
  booktitle={COLT},
  year={2019}
}
We prove a new minimax theorem connecting the worst-case Bayesian regret and minimax regret under partial monitoring with no assumptions on the space of signals or decisions of the adversary. We then generalise the information-theoretic tools of Russo and Van Roy (2016) for proving Bayesian regret bounds and combine them with the minimax theorem to derive minimax regret bounds for various partial monitoring settings. The highlight is a clean analysis of `non-degenerate easy' and `hard' finite… CONTINUE READING

Citations

Publications citing this paper.
SHOWING 1-4 OF 4 CITATIONS

Connections Between Mirror Descent, Thompson Sampling and the Information Ratio

VIEW 10 EXCERPTS
CITES METHODS, BACKGROUND & RESULTS

Differential Privacy for Multi-armed Bandits: What Is It and What Is Its Cost?

VIEW 9 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

Exploration by Optimisation in Partial Monitoring

VIEW 15 EXCERPTS
CITES BACKGROUND, RESULTS & METHODS

References

Publications referenced by this paper.
SHOWING 1-10 OF 36 REFERENCES

Bandit Algorithms

  • T. Lattimore, Cs. Szepesvári
  • 2019
VIEW 6 EXCERPTS
HIGHLY INFLUENTIAL

Further Optimal Regret Bounds for Thompson Sampling

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

Regret Bounds and Minimax Policies under Partial Monitoring

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

Toward a Classification of Finite Partial-Monitoring Games

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

Regret Minimization Under Partial Monitoring

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

Learning to Optimize via Information-Directed Sampling

VIEW 3 EXCERPTS
HIGHLY INFLUENTIAL

Introduction to Nonparametric Estimation

VIEW 1 EXCERPT
HIGHLY INFLUENTIAL