Nonparametric pattern recognition algorithm for testing a hypothesis of the independence of random variables

A new method for testing a hypothesis of the independence multidimensional random variables is proposed. The technique under consideration based on use nonparametric pattern recognition algorithm that meets maximum likelihood criterion. In contrast to traditional formulation problem, there no priori training sample. initial information represented by statistical data, which are made up values multivariate variable. distribution laws in classes estimated according data conditions their dependence and independence. When selecting optimal bandwidths kernel-type probability density estimates, minimum standard deviation used as Estimates error calculated. Based value estimates probabilities errors, decision or variables. developed spectral analysis remote sensing data.