An adaptive backbone formation algorithm for wireless sensor networks
Computer Communications
| Hi-index | 0.01 |
Associative reinforcement learning tasks defined by Barto and Anandan [4] combine elements of problems involving optimization under uncertainty, studied by learning automata theorists, and supervised learning pattern-classification. In our previous work, we presented the SRV algorithm [15] which had been designed for extended versions of associative reinforcement learning tasks wherein the learning system''s outputs could take on real values. In this paper, we state and prove a strong convergence theorem that implies a form of optimal performance (under certain conditions) of the SRV algorithm on these tasks. Simulation results are presented to illustrate the convergence behavior of the algorithm under the conditions of the theorem. The robustness of the algorithm is also demonstrated by simulations in which some of the conditions of the theorem are violated.