Search Techniques in Intelligent Classification Systems

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?