Pattern Classifiers Based on Support Vector Machines

Student: Vyacheslav Glukh

Project Goals  

There already exist many software implementations of Support Vector Machines. The objective of this project is

·  To utilize these implementations in the creation of an environment for conducting experiments by using the Support Vector Machine technique.

·  To extend the functionality of some of existing packages in order to use them as pattern classifiers.

·  To include additional kernels beyond traditional one and study how they influence the pattern classifier performance.

·   To enhance the capabilities of those packages by embedding additional functions (classes) into their source code in order to facilitate the comparison between different kernels and Support Vector Machine implementations and to serve as a platform from which to implement ideas for improving their performance.

·  To perform experiments by using generated data as well as well known databases for pattern recognition.