Block building programming for symbolic regression

Symbolic regression that aims to detect underlying data-driven model, has become increasingly important for industrial data analysis when the experimental model structure is unknown or wrong, or the concerned system has changed. For most of the existing algorithms for symbolic regression, such as genetic programming, the convergence speed might be too slow for large scale problems with a large number of variables. This situation may become even worse with increasing problem size. The aforementioned difficulty make symbolic regression limited in practical applications. Fortunately, in many engineering problems, the independent variables in target models are separable or partially separable. This feature inspires us to develop a new approach, block building programming (BBP), in this paper. BBP divides the original target model into several simple models, and then optimizes them sequentially. Under such circumstance, BBP can make large reductions to the search space. The partition of separability is based on a designed method, block and factor detection. Using two different optimization engines, experiments on a set of symbolic regression problems with separability are conducted. Numerical results show that BBP has a good capability of ‘structure and coefficient optimization’ and a better computational efficiency.