Simulation-Based Methodologies for Global Optimization and Planning
Abstract
The researchers made significant progress in all of the proposed research areas. The first major task in the proposal involved model-based randomized methods for global optimization. In support of this task, the researchers developed new methods for stochastic derivative estimators for discontinuous payoff functions; the method includes Infinitesimal Perturbation Analysis and the Likelihood Ratio method as special cases and can be applied to functions of more general forms containing indicator functions. The researchers developed a new method of distributed ordinal comparison of selecting the best option, which maximizes the average of local reward function values among available options in a dynamic network. They discovered a new innovative approach to simulation-based global optimization by building a connection between global optimization and evolutionary games, as well as another new approach that exploits particle filtering; they have summarized our model-based results in a comprehensive survey paper. The researchers also made significant progress in other model-based randomized methods, including a stochastic search algorithm for solving general optimization problems with little structure; the algorithm iteratively finds high quality solutions by randomly sampling candidate solutions from a parameterized distribution model over the solution space.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 11, 2013
- Accession Number
- ADA591505
Entities
People
- Jiaqiao Hu
- Michael C. Fu
- Steven I Marcus
Organizations
- University of Maryland