Feature selection for high-dimensional remote sensing data by maximum entropy principle based optimization

For high-dimensional remote sensing data, the appropriate selection of features has a significant effect on the cost and accuracy of an automated classifier. In this paper, a method for feature selection by estimation of maximum entropy principle algorithm, is presented. This method based on the EDA (estimation of distribution algorithm) paradigm, avoids the use of crossover and mutation operators to evolve the populations, in contrast to genetic algorithms. It is combined with an approximate application of the maximum entropy principle as the models for representing the probability distribution of a set of candidate solution in the feature selection problem, using the application of automatic learning methods to induce the right distribution model in each generation. Computational comparison is made between EDA in combination with Bayesian networks and EDA in combination with maximum entropy principle. Experiments are performed on an AVIRIS dataset.