While estimation of entropy or mutual information is not identical to estimation of the underlying distribution, a good approach to the distribution estimation problem does help, and is currently the best approach known under some conditions [21]. Results from [13] show promise in using a plug-in estimator based on distribution modeling from patterns.