We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Artificial Intelligence

Title: A double competitive strategy based learning automata algorithm

Authors: Chong Di
Abstract: Learning Automata (LA) are considered as one of the most powerful tools in the field of reinforcement learning. The family of estimator algorithms is proposed to improve the convergence rate of LA and has made great achievements. However, the estimators perform poorly on estimating the reward probabilities of actions in the initial stage of the learning process of LA. In this situation, a lot of rewards would be added to the probabilities of non-optimal actions. Thus, a large number of extra iterations are needed to compensate for these wrong rewards. In order to improve the speed of convergence, we propose a new P-model absorbing learning automaton by utilizing a double competitive strategy which is designed for updating the action probability vector. In this way, the wrong rewards can be corrected instantly. Hence, the proposed Double Competitive Algorithm overcomes the drawbacks of existing estimator algorithms. A refined analysis is presented to show the $\epsilon-optimality$ of the proposed scheme. The extensive experimental results in benchmark environments demonstrate that our proposed learning automata perform more efficiently than the most classic LA $SE_{RI}$ and the current fastest LA $DGCPA^{*}$.
Subjects: Artificial Intelligence (cs.AI)
Cite as: arXiv:1712.00222 [cs.AI]
  (or arXiv:1712.00222v1 [cs.AI] for this version)

Submission history

From: Chong Di [view email]
[v1] Fri, 1 Dec 2017 07:54:53 GMT (983kb)

Link back to: arXiv, form interface, contact.