Adaptive control of constrained finite Markov chains
Automatica (Journal of IFAC)
Simulation-based optimization over discrete sets with noisy constraints
Proceedings of the Winter Simulation Conference
| Hi-index | 22.15 |
The paper considers the problem of minimizing f(x) over the constraint set C = {x: q^i(x) @? 0, i = 1, ..., s}, but the f(.) and q^i(.) are unknown and only noise-perturbed samples of both the f(.) and the q^i(.) are available, at selected parameter settings. Under certain convexity conditions, a stochastic approximation algorithm is set up, and convergence with probability one to the optimum parameter value is proved. Numerous practical examples fit the problem description.