Generalized Maximum Entropy for Supervised Classification
This work addresses classification problems for machine learning practitioners by providing a novel theoretical framework, though it appears incremental as it builds on existing maximum entropy and minimax approaches.
The paper tackles the problem of supervised classification by developing a framework based on the generalized maximum entropy principle, which leads to minimax risk classifiers with performance guarantees through convex optimization, and quantifies their performance compared to conventional methods.
The maximum entropy principle advocates to evaluate events' probabilities using a distribution that maximizes entropy among those that satisfy certain expectations' constraints. Such principle can be generalized for arbitrary decision problems where it corresponds to minimax approaches. This paper establishes a framework for supervised classification based on the generalized maximum entropy principle that leads to minimax risk classifiers (MRCs). We develop learning techniques that determine MRCs for general entropy functions and provide performance guarantees by means of convex optimization. In addition, we describe the relationship of the presented techniques with existing classification methods, and quantify MRCs performance in comparison with the proposed bounds and conventional methods.