COMPARISON OF ENTROPY AND MEAN SQUARE ERROR CRITERIA IN ADAPTIVE SYSTEM TRAINING USING HIGHER ORDER STATISTICS
Deniz Erdogmus and Jose C. Principe
Email: deniz@grove.ufl.edu and principe@cnel.ufl.edu
The errorentropyminimization approach in adaptive
system training is addressed in this paper. The effect of
Parzen windowing on the location of the global minimum
of entropy has been investigated. An analytical proof that
shows the global minimum of the entropy is a local
minimum, possibly the global minimum, of the non
parametrically estimated entropy using Parzen
windowing with Gaussian kernels. The performances of
errorentropyminimization and the meansquareerror
minimization criteria are compared in shortterm
prediction of a chaotic time series. Statistical behavior of
the estimation errors and the higher order central
moments of the time series data and its predictions are
utilized as the comparison criteria.