ESTIMATION OF MUTUAL INFORMATION USING KERNEL DENSITY ESTIMATORS

Mutual information is useful for investigating the dependence between two experimental time series. It is often used to establish an appropr iate time delay in phase-portrait reconstruction from time-series data . A histogram based approach has been used so far to estimate the prob abilities. It is shown here that kernel density estimation of the prob ability density functions needed in estimating the average mutual info rmation across two coordinates can be more effective than the histogra m method of Eraser and Swinney.