Feature Selection in Sensor Environment using a Neural Framework

Speaker:
Prof. Nikhil R. Pal, Indian Statistical Institute, India

Venue: LT14
Date: 3 October 2007, Wednesday
Time: 3.00 - 4.00 pm

Chair: A/Prof Jagath C. Rajapakse

Co-Sponsored by:
IAPR Technical Committee on Pattern Recognition for Bioinformatics
IEEE Computational Intelligence Society Chapter, Singapore

Abstract:
Consider the problem of designing an intelligent weld inspection system. Typically for such problems different sensors such as radiograph, acoustic emission, thermograph, and eddy current are used as sources of information. The raw input from each of the sensors is typically converted into a set of features and used to design the inspection system. One of important problems related to this is to reduce the required number of necessary sensors. This is often demanded for many reasons including cost of design, processing time, cost of decision making and practical utility – for example, a robot equipped with many sensors may not be able to access the desired area because of constraints on the space. Thus, sensor selection becomes a generalized form of feature selection – selecting groups / subsets of features – either reject all features from a sensor (i.e. reject the sensor) or accept all features from a sensor (i.e. accept the sensor).

In this talk, I shall present two neural network based schemes for solving this problem. Both schemes select useful sensors simultaneously with learning of the main decision making task and hence can take into account possible nonlinear interactions between sensors, the problem solving tool and the task at hand. The effectiveness of the systems will be demonstrated using a few examples.

Biodata of Speaker:
Nikhil R. Pal obtained B. Sc. (Physics) and Master of Business Management (Operations research) degrees from the University of Calcutta and M. Tech. (Comp. Sc.) and Ph. D. (Comp. Sc.) degrees from the Indian Statistical Institute, Calcutta. He is a Professor in the Electronics and Communication Sciences Unit of the Indian Statistical Institute, Calcutta.

He is the Editor-in-Chief of IEEE Transactions on Fuzzy Systems. He serves on the editorial /advisory board of several journals including International Journal of Approximate Reasoning, Neural Information Processing - Letters and Reviews, International Journal of Neural Systems, Fuzzy Sets and Systems, and IEEE Transactions on Systems, Man and Cybernetics-B.

He has edited/co-edited several books and co-authored a book titled “Fuzzy Models and Algorithms for Pattern Recognition and Image Processing,” Kluwer Academic Publishers, 1999. His research interest includes bioinformatics, image analysis, pattern recognition, fuzzy sets theory, neural networks, and evolutionary computation.

He has given plenary/keynote speeches in many international conferences. He was the president of Asia Pacific Neural Net Assembly and currently has been serving as a governing board member of the same. He was the Program Chair of the 4th International Conference on Advances in Pattern recognition and Digital Techniques, Dec. 1999, Calcutta, India and the General Chair of 2002 AFSS International Conference on Fuzzy Systems, Calcutta, 2002 and the 11th International Conference on Neural Information Processing, ICONIP 2004. He was a co-program chair of 2005 IEEE International Conference on Fuzzy Systems and 2006 IEEE International Conference on Fuzzy Systems.

He is a Fellow of Indian National Academy of Engineering and a Fellow of the IEEE, USA.