Equation Discovery with Bayesian Tree-Adjoining GrammarsNew

This paper extends tree-adjoining grammar-based nonlinear system identification to a Bayesian formulation for general nonlinear systems. It defines a generative prior over tree structures and their parameters, with candidate models assembled from a finite set of grammatical rules. Unlike existing TAG identifiers that use evolutionary optimization and return point estimates, the proposed approach places model structure within a probabilistic framework.