Identify Discontinuous Parameter of Parabolic System via Point-Tree Structured Genetic Programming

In this paper, we apply point-tree structured genetic programming (PTGP) to identify a discontinuous parameter of parabolic system. The discontinuous parameter can be decomposed into several subcomponents. Each subcomponent can be represented as a simple function. We call these simple functions as sub-functions. We use the PTGP individual to represent the discontinuous parameter, and evolve these individuals to learn it. During the evolution, the subfunctions and their domains co-evolve and co-adapt automatically. The numerical experiments show PTGP can identify the discontinuous parameter successfully. Compare to the stand genetic programming approach, the performance is improved significantly.