 |
|
|
|
|
 
|
Partial List of Tutorials
Application of Neural Network and Cerebellar Model Articulation Controller in Control Problem
| Speaker: |
Professor Chih-Min Lin
Chairman,
Department of Electrical Engineering, Yuan-Ze University, Taiwan |
| Abstract: |
Recently, neural network (NN)-based control technique has represented an alternative method to solve the control problems. The most useful property of neural networks is their ability to approximate arbitrary linear or nonlinear mapping through learning. The basic issues in neural network feedback control are to provide on-line learning algorithms that do not require preliminary off-line tuning. Fuzzy neural network possesses advantages both of fuzzy systems and neural networks since it combines the fuzzy reasoning capability and the neural network on-line learning capability. Recurrent neural network has capabilities superior to feedforward neural networks, such as their dynamic response and their information storing ability. The first part of this talk will introduce neural network, fuzzy neural network and recurrent neural network; and represent their applications in some control problems.
Based on biological prototype of human brain and improved understanding of the functionality of the neurons and the pattern of their interconnections in the brain, a theoretical model used to explain the information-processing characteristics of the cerebellum was developed independently by Marr (1969) and Albus (1971). Cerebellar model articulation controller (CMAC) was first proposed by Albus. CMAC is a learning structure that imitates the organization and functionality of the cerebellum of the human brain. That model revealed the structure and functionality of the various cells and fibers in the cerebellum. The core of CMAC is an associative memory which has the ability to approach complex nonlinear functions. CMAC takes advantage of the input-redundancy by using distributed storage and can learn nonlinear functions extremely quickly due to the on-line adjustment of its system parameters. CMAC is classified as a non-fully connected perceptron-like associative memory network with overlapping receptive-fields. It has good generalization capability and fast learning property and is suitable for on-line application of control systems. This talk also introduces several CMAC-based adaptive control systems; these control systems combine the advantages of CMAC identification, adaptive control and robust control techniques. In these systems, the on-line parameter training methodology, using the gradient descent method and the Lyapunov stability theorem, is proposed to increase the learning capability. Moreover, the applications of these systems in control problems are demonstrated. |
| Biography: |
Prof. Chih-Min Lin is currently a Professor and the Chairman of the Department of Electrical Engineering, Yuan-Ze University, Taiwan. He also serves as the Committee Member of National Science Council, Control Branch; Chinese Automatic Control Society; Taiwan Fuzzy System and Science Society; and Taiwan Systems Engineering Society. During 1986-1992, he was with the Chung Shan Institute of Science and Technology as a Deputy Director of system engineering. He joined the faculty of the Department of Electrical Engineering, Yuan-Ze University, Taiwan, in 1993. During 1997-1998, he was the honor research fellow in the University of Auckland, New Zealand. He has served as the Deputy Chairman of IEEE Control Systems Society, Taipei Chapter in 1999-2000, now he is an IEEE Senior Member.
Prof. Lin’s research interests include fuzzy neural network, cerebellar model articulation controller, control system and systems engineering. He has published 84 journal papers and 120 conference papers. He has been awarded with the outstanding research professor and chair professor. He has given several plenary lectures and invited talks and served as the committee member in several international conferences. Now he also serves as the editorial board of 4 international journals. |
Principles of Stochastic Discrimination and Ensemble Learning
| Speaker: |
Dr. Tin Kam Ho
Mathematical and Algorithmic Sciences Research Center, Bell Labs, USA |
| Abstract: |
Learning in everyday life is often accomplished by making many random guesses and synthesizing the feedback. Kleinberg's analysis of this process resulted in a new method for classifier design -- stochastic discrimination (SD). The method constructs an accurate classifier by combining a large number of very weak discriminators that are generated essentially at random. An important advantage is that classifiers designed by this way are insensitive to overtraining.
SD is an ensemble learning method in an extreme form. Studies on other ensemble methods for classification have long suffered from the difficulty of modeling the complementary strengths of the components. The SD theory addresses this rigorously via mathematical notions of enrichment, uniformity, and projectability.
In this tutorial we explain these concepts via a simple numerical example, with a focus on a fundamental symmetry in point set covering that is the key observation leading to the foundation of the SD theory. We illustrate, step by step, how the SD principle operates in this example. We then describe and discuss more sophisticated implementations for practical uses. We believe a basic understanding of the SD method will open the way to explorations of a new classifier technology, and lead to developments of better tools for analyzing other ensemble methods.
No prior knowledge is assumed other than a general understanding about major concerns in classification. Though, some experience on ensemble methods, such as classifier combination, bagging, boosting, etc. will be helpful in discussions on how the SD method relates to others. |
| Description: |
Click here to download |
| Biography: |
Dr. Ho is a Distinguished Member of Technical Staff in the Mathematical and Algorithmic Sciences Research Center of Bell Laboratories. She received a PhD in Computer Science from State University of New York at Buffalo in 1992. She has published actively on classifier combination, decision forests, stochastic discrimination, data complexity, and many topics in pattern recognition applications. Her 1992 Ph.D. thesis was a pioneering study on classifier combination. She is Editor-In-Chief of the IAPR official journal Pattern Recognition Letters. She has also been on the editorial board of Pattern Recognition (1999-present), International Journal on Document Analysis and Recognition (2003-present), and IEEE-Transactions on PAMI (1999-2002). She is a Fellow of the IAPR and IEEE, and has 7 U.S. patents on classifier design, image analysis, and wireless tracking.
Dr. Ho has been a close collaborator with Prof. Kleinberg on SD since 1992. She also developed several variants of SD implementations, and studied relationship of SD to classifier combination and ensemble learning methodologies. She has published several articles related to SD, on its introduction, experiments, and variants. |
Linguistic models: from data to granular architectures
| Speaker: |
Professor Witold Pedrycz
Department of Electrical & Computer Engineering, University of Alberta, Canada
and
System Research Institute, Polish Academy of Sciences
Warsaw, Poland |
| Abstract: |
Linguistic (granular) models have emerged as a broad category of constructs whose parameters are inherently granular and thus represented in terms of fuzzy sets or rough sets. In panoply of existing models, linguistic models play an important role by striking a sound compromise between model accuracy and transparency while taking into account the quality of data along with their granularity.
The underlying design methodology of linguistic models is aimed at addressing an active and dominant role of the user (designer) in a way that available experimental evidence is looked upon and incorporated into the model from the perspective established by the designer. Such a perspective is reflective of his needs and specific goals of modeling and the required specificity of modeling itself. In a nutshell, the developer of the model can take full advantage of data by casting its processing in the context predetermined by his requirements and the ultimate way in which the model is going to be applied in a given real-world framework.
We elaborate on detailed design algorithms that are used when building linguistic models by guiding the audience through consecutive development steps, offering a thorough analysis and bringing some motivating features as well as pointing at possible alternatives. The granulation of information becomes a crucial design facet of linguistic models. We discuss various methods of fuzzy clustering which are reflective of the modeling perspective (context) assumed by the user. In particular, this concerns a suite of so-called context-based fuzzy clustering techniques.
We also show that some generic versions of linguistic models support a principle of rapid prototyping in which the model is constructed with a minimal learning effort and, if acceptable from the user’s perspective, it could be further refined and augmented. Interestingly, the issue of rapid prototyping has not been raised and discussed in depth when building constructs of Computational Intelligence. As a matter of fact, quite commonly we are engaged in a fairly laborious and computationally demanding design pursuits not being fully confident as to the quality of the model constructed in this way. Given this, in the presentation we discuss essential pros and cons of rapid prototyping in system modeling.
The presentation is self-contained to a significant extent and all essential prerequisites will be covered as a part of this tutorial. |
| Biography: |
Witold Pedrycz is a Professor and Canada Research Chair (CRC - Computational Intelligence) in the Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Canada. He main research directions involve Computational Intelligence, fuzzy modeling, knowledge discovery and data mining, fuzzy control including fuzzy controllers, pattern recognition, knowledge-based neural networks, relational computation, and Software Engineering. He has published numerous papers in this area. He is also an author of 11 research monographs covering various aspects of Computational Intelligence and Software Engineering. Witold Pedrycz has been a member of numerous program committees of IEEE conferences in the area of fuzzy sets and neurocomputing. He currently serves as an Associate Editor of IEEE Transactions on Fuzzy Systems, and a number of other international journals. He is an Editor-in-Chief of Information Sciences and Editor-in-Chief of IEEE Transactions on Systems, Man, and Cybernetics - part A. In 2007 he received a prestigious Norbert Wiener award from the IEEE Systems, Man, and Cybernetics Council. |
Intelligence Pattern Recognition and Applications to Biometrics in an Interactive Environment
| Speaker: |
Professor Patrick Wang
College of Computer and Information Science, Northeastern University, Boston, MA, USA |
| Abstract: |
This talk deals with some fundamental aspects of biometrics and its applications. It basically includes the following subtopics: (1) Overview of Biometric Technology and Applications (2) Importance of Security: A Scenario of Terrorists Attack, (3) What are Biometric Technologies?, (4) Biometrics: Analysis vs Synthesis.(5) Analysis: Interactive Pattern Recognition Concept, (6) Concept of “Semantics” and “Ambiguity”, Their Importance and Applications, (7) How it works: Fingerprint Extraction and Matching, Iris, and Facial Analysis, (8) Authentication Applications, (9) Thermal Imaging: Emotion Recognition. (10) Synthesis in biometrics, (11) Modeling and Simulation, and (12) more Examples and Applications in Interactive Environment. |
| Biography: |
Prof. Patrick Wang, PhD. IAPR Fellow, is Tenured Full Professor, Northeastern University, USA, iCORE (Informatics Circle of Research Excellence) Visiting Professor, University of Calgary, Canada, Otto-Von-Guericke Distinguished Guest Professor, Magdeburg University, Germany, Zijiang Visiting Chair, ECNU, Shanghai, China, as well as honorary advisory professor of several key universities in China, including Sichuan University, Xiamen University, East China Normal University, Shanghai, and Guangxi Normal University, Guilin. Dr. Wang is also IEEE-SMC Outstanding Achievement Award recipient, Harvard Medical, IEEE-BIBE 2007. Prof. Wang received his BSEE from National Chiao Tung University (Jiaotong U.), MSEE from National Taiwan University, MSICS from Georgia Institute of Technology, and PhD, Computer Science from Oregon State University. Dr. Wang has published over 23 books, 130 technical papers, 3 USA/European Patents, in PR/AI/TV/Cybernetics/Imaging, and is currently founding Editor-in-Chief of IJPRAI (International Journal of Pattern Recognition and Artificial Intelligence) , and Book Series of MPAI, WSP. In addition to his technical interests, Dr. Wang also published a prose book, Harvard Meditation Melody mQnand many articles and poems regarding Du Fu and Li Bais poems, Beethoven, Brahms, Mozart and Tchaikovskys symphonies, and Bizet, Verdi, Puccini and Rossinis operas. |
|
|
|
|
|
|
|
|
|
