Jp. Dussault et al., COMBINING THE STOCHASTIC COUNTERPART AND STOCHASTIC-APPROXIMATION METHODS, Discrete event dynamic systems, 7(1), 1997, pp. 5-28
Citations number
36
Categorie Soggetti
Controlo Theory & Cybernetics",Mathematics,"Operatione Research & Management Science",Mathematics,"Operatione Research & Management Science","Robotics & Automatic Control
In this work, we examine how to combine the score function method with
the standard crude Monte Carlo and experimental design approaches, in
order to evaluate the expected performance of a discrete event system
and its associated gradient simultaneously for different scenarios (c
ombinations of parameter values), as well as to optimize the expected
performance with respect to two parameter sets, which represent parame
ters of the underlying probability law (for the system's evolution) an
d parameters of the sample performance measure, respectively. We explo
re how the stochastic approximation and stochastic counterpart methods
can be combined to perform optimization with respect to both sets of
parameters at the same time. We outline three combined algorithms of t
hat form, one sequential and two parallel, and give a convergence proo
f for one of them. We discuss a number of issues related to the implem
entation and convergence of those algorithms, introduce averaging vari
ants, and give numerical illustrations.