Title: Modular Abstract Self-learning Tabu Search

Speaker: Michael Ciarleglio, J. Wesley Barnes, University of Texas at Austin

Date/Time: Monday, April 2, 2007, 10:45am – 11:45am

Location: CSRI Building, Room 90 (Sandia NM)

Brief Abstract: A general-purpose optimization engine, MASTS (Modular Abstract Self-Learning Tabu Search), is based on tabu search, a metaheuristic that relies on memory structures to intelligently navigate the search space of massive complex decision making problems.  The application programming interface (API) that defines how MASTS connects to different problems is flexible and abstract, greatly enabling the widest possible variety of applications.

While under development, MASTS has been applied to several different problems in resource planning.  In groundwater management, MASTS is being used in two different arenas to search for and find significantly superior pumping strategies that satisfy stakeholder requirements, such as minimizing the impact on sensitive environmental features.  In conservation planning for endangered species, MASTS is able to determine ensembles of  efficient and effective designs for conservation area networks that are superior to those constructed by current methodologies using dramatically less computational effort.  This allows the decision maker the freedom to select the most appropriate design based on experience and the context of the problem. The capability and effectiveness of the network depends on the geometric configuration and a variety of economic, social, and ecosystem costs.

The MASTS framework goes well beyond a single optimization to provide an environment where problems can be thoroughly explored.  MASTS contains innovative programming techniques that handle multiple criteria and multiple objectives with ease, assisting users in the design their own searches and providing aid in the analysis of the collective results. 

CSRI POC: Jean-Paul Watson, (505) 845-8887