Physical-based methods for matrix compression in large array problems: a unified view