A unified methodology for categorizing various complexobjects is presented in this book.
Through probability theory novelasymptotically minimax criteria suitable for practical
applications in imagingand data analysis are examined including the special cases such as
theJensen-Shannon divergence and the probabilistic neural network. An optimalapproximate
nearest neighbor search algorithm which allows fasterclassification of databases is featured.
Rough set theory sequential analysisand granular computing are used to improve performance of
the hierarchicalclassifiers. Practical examples in face identification (including deep
neuralnetworks) isolated commands recognition in voice control system andclassification of
visemes captured by the Kinect depth camera are included.This approach creates fast and
accurate search procedures by using exactprobability densities of applied dissimilarity
measures. Thisbook can be used as a guide for independent study and as supplementary
materialfor a technically oriented graduate course in intelligent systems and datamining.
Students and researchers interested in the theoretical and practicalaspects of intelligent
classification systems will find answers to: - Why conventional implementation of the naive
Bayesianapproach does not work well in image classification? - How to deal with insufficient
performance of hierarchicalclassification systems? - Is it possible to prevent an exhaustive
search of thenearest neighbor in a database?
