Parameter adaptive multimodal optimization and its application in smart lighting

Huang, Zhenhua
Thumbnail Image
Other Contributors
Sanderson, A. C. (Arthur C.)
Wen, John T.
Julius, Anak Agung
Mishra, Sandipan
Issue Date
computer , Systems engineering , Electrical
Terms of Use
This electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.
Full Citation
Differential evolution is a class of evolutionary optimization methods that has been successful in a variety of difficult multimodal problem sets. Recently, the JADE algorithm has been developed in our laboratory and provides the capability to adapt algorithm control parameters to a range of different objective function characteristics at different phases of the evolutionary search. JADE has been successful in solving high dimension multimodal optimization problems, and was awarded first place in an international competition. The SPADE algorithm considered in this research is a further refinement of JADE, and introduces the principle of speciation in order to find multiple likely solutions to large multimodal problems. In SPADE, the population evolves into different species corresponding to clusters of candidate solutions. Separate species evolve in parallel and converge to multiple regions of the search space. The speciated populations are guided by the Minimum Description Length (MDL) principle that minimizes the complexity of the data set representation. Adaptive crossover between species further supports flexibility of the search process.
May 2013
School of Engineering
Dept. of Electrical, Computer, and Systems Engineering
Rensselaer Polytechnic Institute, Troy, NY
Rensselaer Theses and Dissertations Online Collection
Restricted to current Rensselaer faculty, staff and students. Access inquiries may be directed to the Rensselaer Libraries.