By David A. Joiner
Kean University, Union, New Jersey
This teaching module introduces stochastic approaches to finding optimum solutions for adequately defined systems. Because these approaches are intrinsically random, large numbers of random samples are typically required to find robust optima. This results in either long single simulation runs, or the need for multiple replicated simulations considered as an ensemble, or both. Monte Carlo, simulated annealing and genetic algorithm approaches to optimization are introduced in this module and applied to a few example problems, and parallelization strategies and their resulting performance gains are assessed.
The module document can be downloaded below and contains details for procuring the source code and documentation.
Stochastic Optimization.docx : This MS Word document introduces deterministic and stochastic methods for optimization, develops the ideas and algorithms for three stochastic optimization approaches, and proposes and tests a parallelization strategy for Monte Carlo optimization.
Stochastic Optimization (.pdf) : The module document in PDF format.