Generalized Maximum Entropy for Supervised Classification

The maximum entropy principle advocates to evaluate events’ probabilities using a distribution that maximizes among those satisfy certain expectations’ constraints. Such can be generalized for arbitrary decision problems where it corresponds minimax approaches. This paper establishes framework supervised classification based on the leads risk classifiers (MRCs). We develop learning techniques determine MRCs general functions and provide performance guarantees by means of convex optimization. In addition, we describe relationship presented with existing methods, quantify in comparison proposed bounds conventional methods.