AN EMPIRICAL DEMONSTRATION OF THE NO FREE LUNCH THEOREM

Abstract

In this paper, we provide a substantial empirical demonstration of the statistical machine learning result known as the No Free Lunch Theorem (NFLT). We specifically compare the predictive performances of a wide variety of machine learning algorithms/methods on a wide variety of qualitatively and quantitatively different datasets. Our research work conclusively demonstrates a great evidence in favor of the NFLT by using an overall ranking of methods and their corresponding learning machines, revealing in effect that none of the learning machines considered predictively outperforms all the other machines on all the widely different datasets analyzed. It is noteworthy however that while evidence from various datasets and methods support the NFLT somewhat emphatically, some learning machines like Random Forest, Adaptive Boosting, and Support Vector Machines (SVM) appear to emerge as methods with the overall tendency to yield predictive performances almost always among the best