Improved Exploration in Factored Average-Reward MDPs
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Documents
- Fulltext
Final published version, 611 KB, PDF document
We consider a regret minimization task under the average-reward criterion in an unknown Factored Markov Decision Process (FMDP). More specifically, we consider an FMDP where the state-action space XX and the state-space SS admit the respective factored forms of X=⊗ni=1XiX=⊗i=1nXi and S=⊗mi=1SiS=⊗i=1mSi, and the transition and reward functions are factored over XX and SS. Assuming a known a factorization structure, we introduce a novel regret minimization strategy inspired by the popular UCRL strategy, called DBN-UCRL, which relies on Bernstein-type confidence sets defined for individual elements of the transition function. We show that for a generic factorization structure, DBN-UCRL achieves a regret bound, whose leading term strictly improves over existing regret bounds in terms of the dependencies on the size of \cSi\cSi’s and the diameter. We further show that when the factorization structure corresponds to the Cartesian product of some base MDPs, the regret of DBN-UCRL is upper bounded by the sum of regret of the base MDPs. We demonstrate, through numerical experiments on standard environments, that DBN-UCRL enjoys a substantially improved regret empirically over existing algorithms that have frequentist regret guarantees.
Original language | English |
---|---|
Title of host publication | Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS) |
Publisher | PMLR |
Publication date | 2021 |
Pages | 3988-3996 |
Publication status | Published - 2021 |
Event | 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021) - San Diego, United States Duration: 13 Apr 2021 → 15 Apr 2021 |
Conference
Conference | 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021) |
---|---|
Land | United States |
By | San Diego |
Periode | 13/04/2021 → 15/04/2021 |
Series | Proceedings of Machine Learning Research |
---|---|
Volume | 130 |
ISSN | 2640-3498 |
- NOISE
Research areas
Links
- https://proceedings.mlr.press/v130/
Final published version
ID: 301365745