Pattern recognition methods as applied to optimize systems represented by simulation models

Abstract

This overview lecture discusses four topics and describes the method of approximate optimization of dynamic stochastic systems known as the optimization simulation method. The grid method of uniform probing of the space of parameters called LP τ search with averaging allows estimating the efficiency region as the region in the space of parameters where the system quality indices are better than in other regions. Ensuring better quality of processing the results of simulation experiments, pattern recognition methods allow speeding up the evaluation of efficiency regions.

Recognition of nonlinear functions for estimating region of efficiency in LP τ-search with averaging

“…In [1,2] we have comprehensively considered the problem of approximate optimization with using a simulation model of a dynamic stochastic object. The solution algorithms and special features of optimiza tion problems statement have been discussed at length in [3,4].…”

“…In other words it is necessary to construct either an estimate of the region of eficiency in which the conditions (2) are met or a set of estimates from the space of esti mates with an assigned metric such that for them qual ity indices form the Pareto set. In the range of integra tion G called the region of eficiency, averaged values of quality indices must be better than values in other ranges.…”