D2.1.2 Mlc ++ D2.1.2.1 Motivation for Mlc ++

MLC++, the Machine Learning library in C++ is a set of libraries and utilities that can aid developers in interfacing machine learning technology, aid users in selecting an appropriate algorithm for a given task, and aid researchers in developing new algorithms, especially hybrid algorithms and multi-strategy algorithms. D2.1.2.1 Motivation for MLC++ A learning algorithm cannot outperform any other learning algorithm when the performance measure is the expected generalization accuracy, assuming all possible target concepts are equally likely. This theoretical result, sometimes called the No Free Lunch Theorem or Conservation Law (Mitchell 1982, Wolpert 1994, Schaffer 1994), implies that it is impossible for any algorithm to have the highest generalization accuracy in all domains. The theorem has few interesting consequences in practice because users are interested in performance on a specific domain or a set of domains, and not all concepts are equally probable. For a given domain, comparisons certainly make sense and some algorithms will perform better than others.