Exploration by Optimisation in Partial Monitoring
Abstract
We provide a simple and efficient algorithm for adversarial $k$action $d$outcome nondegenerate locally observable partial monitoring game for which the $n$round minimax regret is bounded by $6(d+1) k^{3/2} \sqrt{n \log(k)}$, matching the best known informationtheoretic upper bound. The same algorithm also achieves nearoptimal regret for full information, bandit and globally observable games.
 Publication:

arXiv eprints
 Pub Date:
 July 2019
 arXiv:
 arXiv:1907.05772
 Bibcode:
 2019arXiv190705772L
 Keywords:

 Computer Science  Machine Learning;
 Mathematics  Optimization and Control;
 Statistics  Machine Learning
 EPrint:
 high probability bounds, experiments and simplified algorithms/analysis