Two Time-Scale Feasible Direction Method
| Authors: | Tadic Vladislav, University of Sheffield, United Kingdom
Vlajkovic Nenad, University of Belgrade, Yugoslavia
Kyriakopoulos Efstratios, University of Sheffield, United Kingdom |
|---|
| Topic: | 1.4 Stochastic Systems |
|---|
| Session: | Control, Estimation and Analysis of Stochastic Systems |
|---|
| Keywords: | optimization, stochastic approximation, Markov decision problems, Monte Carlo simulation |
|---|
Abstract
Stochastic constrained optimization problems with non-convex objective and convex feasible domainare considered for the case where the objective and constraint functions are available only throuth noisy observations. A general algorithm of the two time-scale stochastic approximation type is proposed for these problems. The proposed algorithm is applied to Markov decision problems with average cost, average constraints and parameterized randomized policy. The asymptotic behavior of the proposed algorithm is analyzed for the case where the algorithm step-sizes are constant and the noise in the observations of the objective and constraint functions depends on the algorithm iterates.
