Keynote Speech【2】

Ramesh Kumar Agrawal


Dean and Professor of School of Computer and Systems Sciences, Jawaharlal Nehru University, Delhi, INDIA


Ramesh Kumar Agrawal is currently Dean and Professor of School of Computer and Systems Sciences, Jawaharlal Nehru University, Delhi, INDIA. He has published more than 40 articles in journals such as Pattern Recognition, Pattern Recognition letter, Neural Computing and Applications, Expert Systems with Applications, International journal of imaging systems and technology, Artificial Intelligence Review, Computing and Informatics, Physical Review and Informatics. He has also published 45 articles in international conferences.
He received his M. Sc and Ph. D from University of Delhi. He did his M. Tech from Indian Institute of Technology, Delhi. He was awarded Junior Research Fellowship and Senior Research Fellowship to obtain PhD at University of Delhi from CSIR, INDIA. He has also worked in University of Delhi, Delhi, India. He has been widely involved in the organization of over ten international conferences and delivered invited lectures in many conferences and workshops. His research interests include pattern recognition, data mining, medical imaging and information security. He supervised eleven PhD and 26 Master students. He teaches Pattern Classification, Data Mining, Image Processing and Algorithm Design and Analysis.

Topic: Optimal Decision Tree Based Multi-class Support Vector Machine

Support Vector Machine (SVM) has been proved to be a successful learning machine in literature, especially for classification. Since SVM was originally designed for binary classification, it is not easy to extend binary SVM to multi-class problem. Constructing k-class SVMs (k > 2) is an on-going research issue. Two approaches are suggested in literature to solve multi-class SVM. One is considering all data in one optimization. The other is decomposing multi-class into a series of binary SVMs, such as "One-Against-All" (OAA) and "Oneversus-One" (OvO). It has been reported in literature that both conventional OvO and OAA SVMs suffer from the problem of unclassifiable region. To resolve unclassifiable region in conventional OvO, we propose decision tree OvO SVM formulation to solve multi-class problems. To maintain high generalization ability, the optimal structure of decision tree is determined using statistical measures for obtaining class separability. The proposed optimal decision tree SVM (ODT-SVM) takes advantage of both the efficient computation of the decision tree architecture and the high classification accuracy of SVM. Experimental results and statistical tests have shown that the proposed ODT-SVM is significantly better in comparison to conventional OvO and OAA in terms of both training and testing time