Symbolic regression via genetic programming to derive empirical models and scaling laws as monomial or polynomial expansions