A new formulation for symbolic regression to identify physico-chemical laws from experimental data

摘要: Abstract A modification to the mixed-integer nonlinear programming (MINLP) formulation for symbolic regression was proposed with aim of identification physical models from noisy experimental data. In formulation, a binary tree in which equations are represented as directed, acyclic graphs, is fully constructed pre-defined number layers. The introduced results reduction required variables and removal redundancy due possible symmetry formulation. tested using numerical found be more efficient than previous literature example respect numbers predictor training data points. globally optimal search extended identify cope noise variable. methodology proven successful identifying correct describing relationship between shear stress rate both Newtonian non-Newtonian fluids, simple kinetic laws chemical reactions. Future work will focus on addressing limitations present solver enable extension target problems larger, complex models.