Keynote Speeches and Plenary Talks

December 11, 2012, Tuesday (Venue: Cinnamon Room)

Zongben Xu, Xi'an Jiaotong University, China

8:40am - 9:35am

Is Extreme Learning Machine Feasible? A Theoretical Assessment
Abstract An extreme learning machine (ELM) is a feed-forward neural network (FNN) alike learning system whose connections between hidden layers and output neurons are adjustable, while hidden layers (the connections from input layers to hidden layers and hidden layers to hidden layers, and hidden neuron biases) can be randomly generated. Numerous applications have demonstrated the feasibility and high efficiency of ELM. It is, however, still an open problem if the feasibility of ELM remains true for any general application. In this talk we provide an answer to this long-standing problem by theoretically justifying the followings: (i) ELM can attain the theoretical generalization bound of the conventional FNNs (in which all connections are adjusted), i.e., ELM does not degrade the generalization capability of the FNNs even when the hidden layers are randomly fixed; (ii) the number of hidden neurons needed for an ELM to achieve the theoretical bound can be estimated (thus providing a theoretical guide to the choice of ELM application); (iii) whenever the activation function is taken as polynomials, the deduced hidden layer output matrix is of full column-ranked, therefore the generalized inverse technique can be applied to yield the solution of ELM. The obtained theoretical results not only justify the feasibility and efficiency of ELM, but also yield various generalizations and improvements of ELM. Simulation results are provided to show the correctness of the obtained theoretical results.
Zongben Xu Zongben Xu was born in 1955. He received his Ph.D. degrees in mathematics from Xi'an Jiaotong University, China, in 1987. He now serves as Vice President of Xi'an Jiaotong University, the Chief Scientist of National Basic Research Program of China (973 Project), and Director of the Institute for Information and System Sciences of the university. He is owner of the National Natural Science Award of China in 2007, and winner of CSIAM Su Buchin Applied Mathematics Prize in 2008. He delivered a 45 minute talk in the International Congress of Mathematicians 2010. He was elected as a member of Chinese Academy of Science in 2011. His current research interests include intelligent information processing and applied mathematics.

Jonathan Wu, University of Windsor, Canada

9:35am - 10:30am

Analytic Training Approach for Object/Action Recognition
Abstract The computer vision community is faced with the challenge of devising novel, robust and efficient algorithms to learn models which are helpful in categorizing huge amount of visual data. Recognition algorithms play a pivotal role in commercially available frameworks for automated analysis and detection of objects of interest (OI). Traditionally, supervised learning frameworks have inherent limitations of longer durations for training and finding local maxima that may lead to poor classification accuracy. Recent action recognition schemes have ignored complexity associated with redundant training samples and learning strategies. To cope with inherent limitations of gradient descent approach, model learning can be analytically performed at an extremely fast speed without iterative adjustments (i.e., extreme learning machine (ELM) and its online sequential variant (OS-ELM)). This presentation is mainly concentrated on recent trends in action and activity recognition. We believe that computation, and selection of meaningful features and their use in model learning are equally important for recognition purposes. In this presentation, first we will discuss reasons for analytic training approach and then present a general framework to efficiently identify OI in still images and later extend its application to human action recognition in videos as well. Such scheme can also be implemented in a situation where training data is coming in a serial mode and training needs to be performed in an incremental fashion.
Jonathan Wu Jonathan Wu is a Professor of Electrical and Computer Engineering and a Tier 1 Canada Research Chair in Automotive Sensors and Information Systems since 2005. He is the founding director of the Computer Vision and Sensing Systems Laboratory at the University of Windsor, Canada. Prior to joining the University, Dr. Wu was a Senior Research Officer and Group Leader at the National Research Council of Canada (NRC). He has published one book in the area of 3D vision and more than 250 peer-reviewed papers (including 90 journal publications) in areas of computer vision, multimedia information processing, and intelligent systems. Dr. Wu is an Associate Editor for IEEE Transaction on Systems, Man, and Cybernetics (part A). Dr. Wu has served on the Technical Program Committees and International Advisory Committees for many prestigious conferences including IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

December 12, 2012, Wednesday (Venue: Cinnamon Room)

Xi-Zhao Wang, Hebei University, China

8:40am - 9:35am

Architecture Selection of ELMs and Their Improved Training Algorithms
Abstract How to automatically determine the architectures of neural networks including Extreme Learning Machines (ELMs) is a big challenge in machine learning and computational intelligence communities. This talk presents a new approach to the automatic selection of ELM architectures based on our proposed Locally Generalized Error Model (LGEM). An error upper bound which includes the training error and sensitivity is given for ELMs. An learning algorithm with accelerated training speed is designed which improves the computation of the M-P inverse matrix. Experimental results have shown that the LGEM model and the accelerating training algorithm are promising in applications.
Xi-Zhao Wang Xi-Zhao Wang is presently the Dean and Professor of the College of Mathematics and Computer Science, Hebei University, China. He received his Ph.D. degree in Computer Science from Harbin Institute of Technology, Harbin, China, in 1998. From September 1998 to September 2001, he served as a Research Fellow in the Department of Computing, Hong Kong Polytechnic University, Hong Kong. He became Full Professor and Dean of the College of Mathematics and Computer Science in Hebei University in October 2001. His main research interests include learning from examples with fuzzy representation, fuzzy measures and integrals, neuro-fuzzy systems and genetic algorithms, feature extraction, multi-classifier fusion, and applications of machine learning. He has 160+ publications including 6 books, 7 book chapters, and 100+ journal papers in IEEE Transactions on PAMI/SMC/FS, Fuzzy Sets and Systems, Pattern Recognition, etc. He has been the PI/Co-PI for more than 20 research projects supported partially by the National Natural Science Foundation of China and the Research Grant Committee of Hong Kong Government. Prof Wang is the Editor-in-Chief of International Journal of Machine Learning and Cybernetics. He is a member of IEEE SMC Society Board of Governor member in 2005, 2007-2009, 2012-2014; the Chair of IEEE SMC Technical Committee on Computational Intelligence, an Associate Editor of IEEE Transactions on SMC, Part B; an Associate Editor of Pattern Recognition and Artificial Intelligence; an Associate Editor of Information Sciences; an executive member of Chinese Association of Artificial Intelligence; and an executive member of Chinese Industrial and Applied Mathematics. Prof. Wang was the recipient of the IEEE-SMCS Outstanding Contribution Award in 2004 and the recipient of IEEE-SMCS Best Associate Editor Award in 2006. Also he is the recipient of 2008 IEEE Outstanding SMCS Chapter Award and 2009 Most Active SMC Technical Committee Award. He is the General Co-Chair of the 2002-2012 International Conferences on Machine Learning and Cybernetics, co-sponsored by IEEE SMCS. He is a Fellow of IEEE.

Donald C. Wunsch, Missouri University of Science & Technology, USA

9:35am - 10:30am

Blurring the Distinctions Between Supervised and Unsupervised learning
Abstract Clustering, or unsupervised learning, is receiving increasing attention due to the rapidly expanding capability of society to create data. Often this data must be analyzed by unsupervised approaches for data-driven or computational complexity reasons. However, even in these situations, some ground truth may be available, similar to what is used in the supervised case. Furthermore, techniques for simplifying and preprocessing the data can further blur the distinction between supervised and unsupervised approaches. Extreme Learning Machines (ELM) are a case in point. The presentation will explore these principles including examples from the ELM class of approaches.
Donald C. Wunsch Donald C. Wunsch is the M.K... Finley Missouri Distinguished Professor at Missouri University of Science & Technology (Missouri S&T). Earlier employers were: Texas Tech University, Boeing, Rockwell International, and International Laser Systems. His education includes: Executive MBA - Washington University in St. Louis, Ph.D., Electrical Engineering - University of Washington (Seattle), M.S., Applied Mathematics (same institution), B.S., Applied Mathematics - University of New Mexico. Key research contributions are: Clustering; Adaptive resonance and Reinforcement Learning architectures, hardware and applications; Neurofuzzy regression; Traveling Salesman Problem heuristics; Robotic Swarms; and Bioinformatics. He is an IEEE Fellow and previous INNS President, INNS Fellow and Senior Fellow 07 ¨C present, and served as IJCNN General Chair, and on several Boards, including the St. PatrickĄ¯s School Board, IEEE Neural Networks Council, International Neural Networks Society, and the University of Missouri Bioinformatics Consortium. He chairs the Missouri S&T Information Technology and Computing Committee, a Faculty Senate Standing Committee. He has produced 16 Ph.D. recipients in Computer Engineering, Electrical Engineering, and Computer Science; has attracted over $8 million in sponsored research; and has over 300 publications including nine books. His research has been cited nearly 6000 times.

Plenary Talk (December 12th, 2012)

Narasimhan Sundararajan, Nanyang Technological University, Singapore

5:20pm - 5:45pm

Fast Learning Circular Complex-Valued Extreme Learning Machine (CC-ELM) for Real-Valued Classification Problems
Abstract In this talk, we present a fast learning fully complex-valued extreme learning machine classifier, referred to as Circular Complex-valued Extreme Learning Machine (CC-ELM) for handling real-valued classification problems. CC-ELM is a single hidden layer network with non-linear input and hidden layers and a linear output layer. A circular transformation with a translational/rotational bias term that performs a unique one-to-one transformation of the real-valued feature to the Complex plane is used as an activation function for the input neurons. The hidden layer neurons employ a fully complex-valued Gaussian-like activation function. The input parameters of the CC-ELM are chosen randomly and the output weights are computed analytically. Performance of the CC-ELM is evaluated using a set of benchmark real classification problems from the University of California, Irvine machine learning repository. Finally, the performance of the CC-ELM is compared with existing methods on two real world problems, viz., an acoustic emission signal classification problem and a mammogram classification problem. These study results show that the CC-ELM performs better than other existing (both) real-valued and complex-valued classifiers, especially when the data sets are highly unbalanced.
Narasimhan Sundararajan Narasimhan Sundararajan received the B.E in Electrical Engineering with First Class Honors from the Alagappa Chettiar College of Engg. and Tech., Karaikudi, University of Madras in 1966, M.Tech from the Indian Institute of Technology, Madras in 1968 and Ph.D. In Electrical Engineering from the University of Illinois, Urbana-Champaign in 1971. From 1972 to 1991, he was working in the Indian Space Research Organization, Trivandrum, India starting as a Control System Designer to Director, Launch Vehicle Design Group contributing to the design and development of the Indian satellite launch vehicles SLV3, ASLV, PSLV and GSLV. He worked as the Project Engineer (Mission) for the first Indian Satellite Launch Vehicle project SLV3 team working directly under Dr. Kalam. He was also a NRC Research Associate at NASA - Ames in 1974 and a Senior NRC Research Associate at NASA Langley in 1981-86 under the National Academy of Sciences, USA program. From February 1991, he was working in the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, as a Professor and has retired from that position in July 2010. Currently, he is an emeritus professor visiting Indian Institute of Space Science and Technology, Trivandrum, India. He was a Prof. I.G.Sarma Memorial ARDB Professor (an endowed visiting professor) during Nov.2002 - Feb 2003, at the School of Computer Science and Automation, Indian Institute of Science, Bangalore, India. His research interests are in the areas of aerospace control, machine learning, neural networks and applications and computational intelligence and have more than 250 papers and also five books in the area of neural networks. Dr. Sundararajan is a Fellow of IEEE, an Associate Fellow of AIAA and also a Fellow of the Institution of Engineers, (IES) Singapore. He was an Associate Editor for IEEE Trans. on Control Systems Technology, IFAC Journal on Control Engineering Practice (CEP), IEEE Robotics and Automation Magazine and for Control - Theory and Advanced Technology (C-TAT), Japan. He was also a member of the Board of Governors (BoG) for the IEEE Control System Society (CSS) for 2005. He has contributed as a program committee member in a number of international conferences and was the General Chairman for the Sixth Intl. Conf. On Automation, Robotics, Control and Computer Vision - ICARCV2000 held in Singapore on Dec. 2000.