empirical risk minimization

in computer science, a way to determine theoretical bounds in machine learning