Maximum Entropy Density Estimation Using a Genetic Algorithm

Several unsupervised learning algorithms, neural networks, and support vector machine based classification and clustering approaches are kernel-based, and require sophisticated algorithms for density estimation. The density estimation problem is a nontrivial optimization problem and most of the existing density estimation algorithms provide locally optimal solutions. In this paper we use an entropy maximizing approach that uses global search genetic algorithm to estimate densities for a given data set. Unlike the traditional local search approaches, our approach uses global search and is more likely to provide solutions that are close to global optimum. Using a simulated dataset, we compare the results of our approach with the maximum likelihood approach.