WCCI 2006 will feature a number of pre-congress tutorials covering fundamental and advanced computational intelligence topics. Tutorial proposals, submitted to Tutorials Chair via emails, are solicited and should include title, outline, expected enrollment, and presenter biography. Any inquires regarding the tutorials should address to Tutorial Chair: DeLiang Wang at dwang<@>cse.ohio-state.edu. The deadline for tutorial proposal is January 31, 2006.

Approved Tutorial Workshops

Tutorial AM-1, July 16, 2006 (Sunday), 8:30am-11:00am, Junior Ballroom CD
Principles and Applications of Neural Networks

Scope
The objective of this tutorial is to review the basic ideas of adaptivity as used in adaptive filters and in feedforward neural networks. Applications to nonlinear signal processing and control will be presented. Gradient algorithms, LMS for adaptive filters and backpropagation for neural networks, will be explained and compared. Speed of convergence, misadjustment, and stability will be discussed. New and unusual uses of feedforward neural networks in the development of artificial "cognitive memory" and applications of this memory to pattern recognition and control will be described.

Tutorial MM-1, July 16, 2006 (Sunday), 12:30pm-3:00pm, Junior Ballroom CD
Predictive Learning and Philosophy of Science

Scope
MOTIVATION: The field of Predictive Learning (aka Soft Computing, Machine Learning, Pattern Recognition and Data Mining) is concerned with estimating ‘good' or useful models from available data. Such problems can be usually stated in the framework of inductive learning, where the goal is to come up with a model (generalization) from several known facts (data samples). Since the number of data samples is finite, such problems are usually ill-posed, and this leads to the development of numerous learning algorithms. The main methodological problem, however, is that there is little agreement on the main concepts and issues underlying these algorithms, so the field remains highly fragmented.

The classical philosophy of science is also concerned with principles of inductive learning and other epistemological issues. So there is a clear connection between predictive learning and the philosophy of science. The goal of this tutorial is to investigate this connection, and to relate the main concepts developed in Vapnik-Chervonenkis (VC) learning theory to similar concepts and principles in the philosophy of science. The presented material will be based, to a large extent, on the pioneering ideas introduced by Vapnik [Vapnik, 1998, 2006].

CONTENT: Presentation will start with general background on the framework and main concepts in VC-theory and the philosophy of science. The first part of this tutorial will explore in detail the connection between the VC-dimension and two philosophical principles, parsimony (Occam's razor) and Popper's falsifiability. In the second part , we discuss a novel interpretation of the concept of margin (in SVM methods) using Popper's falsifiability [Cherkassky and Ma, 2006]. This interpretation leads to a very general view of many margin-based learning formulations, including SVM classification, SVM regression, SVM novelty detection etc. In the third part , we discuss several non-standard (non-inductive) learning formulations and new types of inference recently proposed by Vapnik (2006). These new formulations can be motivated by the general philosophical concept of falsifiability. Throughout this tutorial, many important points will be illustrated by empirical comparisons and related to practical applications (fMRI data analysis etc.)

INTENDED AUDIENCE: Researchers and practitioners in machine learning who interested in understanding of the main principles underlying various learning algorithms, and their connection to the philosophy of science. This tutorial is also helpful for developing improved understanding of methodological issues related to application and comparison of different learning methods to real data.

References
Cherkassky, V. and Y. Ma, Data complexity, margin-based learning and Popper's philosophy of inductive learning, invited paper in Data Complexity in Pattern Recognition , M. Basu and T. Ho , Eds, Springer, 2005 (to appear).

Vapnik, V., Statistical Learning Theory, Wiley, 1995.

Vapnik, V., Estimation of Dependencies Based on Empirical Data, Second Edition, Springer 2006 (to appear).

Tutorial PM-1, July 16, 2006 (Sunday), 3:30pm-6:00pm, Junior Ballroom CD
Scalable Clustering Techniques for Data Mining

Scope
Clustering or segmentation has been wide studied across multiple disciplines for over 40 years, leading to a vast array of techniques. However several data mining applications involve complex data characteristics or domain constraints that severely challenge such classical techniques. I will provide a review of recent next-generation clustering methods that address some of these challenges, including methods for (i) clustering very high dimensional directional data such as text and gene expressions, (ii) partial and overlapping clustering, for example to identify genes that fully participate in more than one process (cluster), (iii) market-basket analysis of very larger detail data, (iv) scalable, balanced clustering of a variety of data types including sequences and (v) obtaining a consensus among several multi-view clustering solutions. Applications that naturally lead to these various scenarios will be identified. Several of the solutions described have received best paper awards at top data mining conferences.

Schedule:
1. Data mining challenges
Illustration: Clustering high-dimensional market baskets; Chameleon

2. Generalizing K-means (hard and soft) to cater to a large variety of data types

3. Understanding Generative Approaches to Clustering Complex Data (sequences, graphs, etc)

4. Distributed Clustering with Privacy Preservation

5. Co-clustering with Application

6. Introduction to Semi-supervised Clustering

Tutorial AM-2, July 16, 2006 (Sunday), 8:30am-11:00am, Junior Ballroom AB
Fuzzy Sets and Pattern Recognition

Scope
Hour 1: Fuzzy Sets 101
Topics: Uncertainty and fuzzy models, membership functions, basic fuzzy sets operations, questions about fuzzy sets

Hour 2: Pattern Recognition 101
Topics: Numerical pattern recognition, feature analysis, clustering and classifier design

Tutorial MM-2, July 16, 2006 (Sunday), 12:30pm-3:00pm, Junior Ballroom AB
Evolvable Neural-, Fuzzy-, and Hybrid Systems: Methods and Applications

Scope
Evolvable intelligent systems (EIS) are systems that develop their structure, their functionality and their internal knowledge representation through continuous learning from data and interaction with the environment. EIS can also evolve through generations of populations using evolutionary computation, but the focus of the tutorial is on the adaptive learning and improvement of each individual system. The learning process can be: on-line, off-line, incremental, supervised, unsupervised, active, sleep/dream, etc. These general principles can be applied to develop different models of computational intelligence. The tutorial presents in its first part the theory and the methods of evolving connectionist systems – neural networks that evolve neurons and connections in an incremental way to capture structures and relationship from a stream of data [1], evolving rule based and fuzzy systems (EFS) [2], and some integrated, hybrid models.

The second part of the tutorial presents and demonstrates, using specialised software, numerous practical applications of EIS in: bioinformatics, neuroinformatics, neuro-genetics, autonomous robot control, adaptive speech recognition for mobile devices, process and plant control, car emission control and environmental modelling, evolvable chip design.

The tutorial targets computer scientists, neuroscientists, biologists, engineers and graduate students.

[1] N.Kasabov, Evolving connectionist systems: Methods and Applications in Bioinformatics, Brain study and intelligent machines, Springer, 2002.
[2] P.Angelov, Evolving rule-based systems, Springer Verlag, Physica Verlag, 2002.

Tutorial PM-2, July 16, 2006 (Sunday), 3:30pm-6:00pm, Junior Ballroom AB
Fuzzy Reinforcement Learning

Scope
Following successful tutorial presentations on Fuzzy Reinforcement Learning at Fuzz-IEEE in Barcelona, Spain and at FUZZ-IEEE in Seoul, Korea, Dr. Berenji proposes to teach this tutorial at FUZZ-IEEE in Vancouver. Reinforcement Learning is an important learning technique for learning from interactions with the environment. It is claimed that all future intelligent systems must be able to learn from the environment and this Congress is a very proper setting for a tutorial on this topic. Fuzzy Reinforcement Learning (FRL) extends Reinforcement Learning and enables it to deal with Fuzzy Systems. In this tutorial, we will first discuss Reinforcement Learning and then the main ways that the future fuzzy systems can learn by RL to improve their performances. In particular, here is the outline of this tutorial.

Schedule:
1. Reinforcement Learning
- What is it?
- Evaluative Feedback
- Dynamic Programming
- Monte Carlo Methods
- Temporal Difference Learning
- Generalization and Function Approximation

2. Fuzzy Reinforcement Learning
- Generalized Approximate Reasoning based Intelligent Control (GARIC)
- Fuzzy Q-Learning
- Actor Critic based Fuzzy Reinforcement Learning (ACFRL)

3. Case Studies
- Cart Pole Balancing
- Wireless Battery Power Control
- Internet Node distribution

The expected participants of this tutorial will be engineers and educators interested in developing future intelligent systems with real-time learning capabilities. Based on the presenter past experience in teaching tutorials at FUZZ-IEEE conferences, it is expected that this tutorial will be attended by at least 30 participants.

Tutorial AM-3, July 16, 2006 (Sunday), 8:30am-11:00am, Pavilion Ballroom CD
Evolutionary Computation: A Unified Approach

Scope
The field of Evolutionary Computation (EC) has experienced tremendous growth over the past 15 years, resulting in a wide variety of evolutionary algorithms and applications. The result poses an interesting dilemma for many practitioners in the sense that, with such a wide variety of algorithms and approaches, it is often hard to se the relationships between them, assess strengths and weaknesses, and make good choices for new application areas.This tutorial is intended to give an overview of EC via a general framework that can help compare and contrast approaches, encourage crossbreeding, and facilitate intelligent design choices. The use of this framework is then illustrated by showing how traditional EAs can be compared and contrasted with it, and how new EAs can be effectively designed using it.Finally, the framework is used to identify some important open issues that need further research.

Tutorial MM-3, July 16, 2006 (Sunday), 12:30pm-3:00pm, Pavilion Ballroom CD
Evolutionary Robotics

Scope
Evolutionary Robotics is the artificial evolution of control systems (mainly neural networks) for robots that display autonomous behaviour. The tutorial will describe the methodology used in Evolutionary Robotics, show how it can generate compact and yet powerful control systems, and describe examples that address open questions in biology, neurophysiology, and cognitive science.

Topic: Motivations for Evolutionary Robotics from Biology, Cognitive Science, and Engineering. Basic methodology. Example: evolution of simple navigation skills. The principle of parsimony: Evolution of complex behaviours using simple reactive mechanisms. The role of fitness functions. Evolution of vision-based navigation. The role of the environment in the evolutionary process and the emergence of spatial maps. Co-evolution of active vision and feature sensitivity. Evolution and learning. The role of environmental change. Evolution of simple adaptive rules and a new perspective on learning. Competitive co-evolution in biology, robotics, and humanmachine interaction. The case of predator-prey robots. Introducing learning in coevolutionary systems. Co-evolution of cooperation among robotic and biological organisms. Evolution of spiking control networks and other exotic neural controllers. Evolution of robot morphologies: methods and results.

The topics will be illustrated with examples on a large variety of robots with wheels, legs, and wings. Several experiments described in the tutorial can be replicated by freely available software published and maintained on the lab web page: http://lis.epfl.ch

Literature: Nolfi, S. & Floreano, D. Evolutionary Robotics. The Biology, Intelligence,and Technology of Self-Organizing Machines. MIT Press, 2000-2004.

Tutorial PM-3, July 16, 2006 (Sunday), 3:30pm-6:00pm, Pavilion Ballroom CD
Evolutionary Multi-Objective Optimization

Scope
Many real-world search and optimization problems from sciences, engineering, and commerce are naturally posed as mathematical programming problems involving multiple objectives. The main difficulty in handling multiple conflicting objectives is that they result in a number of optimal solutions (known as Pareto-optimal solutions), instead of a single optimum. Due to the lack of suitable techniques for finding multiple optimal solutions by classical means, such problems are artificially converted into a single-objective optimization problem and solved. Unfortunately, the outcome of such methods is quite dependent on the adopted conversion procedure. In the recent past, evolutionary algorithms are proposed to solve these problems in a less-subjective and efficient manner. Instead of finding one solution at a time, evolutionary multi-objective optimization (EMO) methods find a number of Pareto-optimal solutions in one simulation and leave the decision-making task for later. Such techniques are increasingly getting popular in practice mainly because of three reasons: (i) a range of optimal solutions allows a better decision-making (ii) optimal solutions reveal salient insights about the problem, a matter which has a deeper implications beyond simply finding the optimal solutions, and (iii) many other optimization problems can be better solved by posing them as a multi-objective optimization problem.

In this tutorial, besides introducing the basic concepts of multi-objective evolutionary optimization, a brief review of the state-of-the-art techniques practiced in this emerging field will be discussed. A number of real-world case studies will be shown to clearly demonstrate the advantages of using EMO over classical methods. The research and application in EMO are in their peaks and offer numerous scopes for future investigations -- computationally faster EMO, parallel EMO, EMO for approximate optimal front, interactive EMO, hybrid EMO-Classical methodologies, robust and reliability-based EMO, EMO with complexity estimate, EMO test problem generator, EMO with a large (20+) number of objectives, etc. will be covered. These topics should provide new directions for research to beginners of the EMO area and also to experienced EMO researchers. For applicationists, potential application domains of EMO and key algorithms and their availability in terms of free/commercial softwares will be discussed. A new and exciting by-product of the use of multi-objective optimization is the act of `innovization', in which innovative design principles are deciphered in real-world problems. This activity goes beyond finding an optimal solution and provides useful insights about the `recipe' of solving the problem at hand - a matter which cannot be achieved by any other means. At the end, the importance of developing interactive EMO methodologies involving classical multi-criterion decision-making methods and EMO in a synergistic manner will be highlighted.

In short, this tutorial will introduce the field of evolutionary multi-objective optimization, present the philosophies behind the approaches, contrast them with their classical counterparts, discuss emerging and futuristic research issues, illustrate real-world case studies, and show a number of interesting and unparallel advantages of solving a problem in a multi-objective manner.

Expected enrollment: Due to the popularity and usefulness of EMO procedures, I anticipate a large number of participants for this tutorial. Since ANN and Fuzzy groups will also be there in WCCI-2006, I expect many participants to attend this seminar from these groups as well.

Tutorial AM-4, July 16, 2006 (Sunday), 8:30am-11:00am, Pavilion Ballroom AB
Towards an Autonomous Computationally Intelligent System

Scope
The following points will be covered in the tutorial:
1. The need for brain-guidance in building a safe autonomous computational intelligent system, in terms of societal management;

2. Overview of the EPSRC UK Attention Learning Project (started Feb, 2004), the GNOSYS EC Project (started Oct, 2004) on creating a cognitive robot and the associated MATHESIS project (started Feb, 2006) on creating a robot-based motor learning system;

3. Results on the GNOSYS percept/concept hierarchical modular visual system, and its relation to developments in machine vision;

4. The attention control system in GNOSYS and MATHESIS and their basis in the brain;

5. Overview of insertion of value maps and drives to lead to emotion guidance, and their relation to emotions in the brain;

6. The development of reasoning systems under attention control (involving both visual and motor attention control as well as internal forward and inverse models) and its relation to data on animal learning/creativity experiments;

7. Present results of the projects;

8. Conclusions

Tutorial MM-4, July 16, 2006 (Sunday), 12:30pm-3:00pm, Pavilion Ballroom AB
Biologically Motivated Mental Architectures

Scope
The fields of neural science and psychology have generated very rich results about how the brain works. There is no lack of computational models that characterize the computational architecture of the brain at different degrees of detail. However, there is a lack of architecture models that incorporate perceptual signal processing, cognitive (logic) processing, and the developer of the processors. Understanding properties of mental architectures is of fundamental importance for understanding neural processing systems and their adaptation and learning.

This tutorial systematically reviews key properties of mental architecture using existing major studies and models as examples, including perceptual architectures (e.g., Neisser's two stage visual processing scheme; Feldman & Ballard's 100-step rule; John Tsotsos' model of immediate vision, HMM models ), cognitive architectures (e.g., Soar proposed by Laird, Newell & Rosenbloom, ACT-R by Anderson, and the architecture outline by Albus), motor architectures (e.g., the subsumption architecture by Rodney Brooks and others), value system architectures (e.g., reinforcement learning, Q-learning, and other recent more complete models). Based on recent results from neural science, psychology and computational intelligence, this tutorial further explain a series of properties for higher biological mental architectures, including nonassociative and associative learnabilities, attention selectability, rehearsability, self-awarness, self-effectiveness, observation drivability, developability, abstractability, along with the corresponding architecture components. Architecture examples are used to illustrate such architecture properties. The series of architectural theory explains how a neural system that does not contain any pre-defined symbolic internal representation can be autonomously developed (through programming, prenatal growth and postnatal experience from environment) to deal with not only perceptual tasks such as recognition and classification, but also sensor-driven higher cognitive tasks, such as abstraction, logical reasoning, thinking, planning, and language acquisition and understanding.

Tutorial topics:
1. Review of the history of biologically motivated architectures
2. Biological mental development
3. Review of animal learning theories, nonassociative and associative learning, classical conditioning, instrumental conditioning, time sequence learning, cognitive learning
4. Supervised, reinforcement, communicative learning, and the refined 8 learning types
5. Perceptual processing: retina, LGN, visual cortex as visual processing examples
6. Cognitive processing: recognition, classification, and invariance
7. Motor processing: rehearsal and coupling of effectors
8. Value system: limbic system, intention, value, and their development
9. A hierarchy of mental architectures: non-observation-driven, observation-driven, attention selective, rehearsable, self-aware and self-effecting, developmental, multi-level
10. Example architectures and experimental studies

Length: A half day (3 to 4 hours)

Prerequisites: General programming experience, basic knowledge about vector and matrix operations.

Audience: researchers in biological neural systems, signal processing, image processing, computer vision, pattern recognition, speech recognition, autonomous navigation, autonomous control, language processing, robotics, human-machine interface, and artificial intelligence.

Handout: Tutorial material will be provided by the host institution.

Tutorial PM-4, July 16, 2006 (Sunday), 3:30pm-6:00pm, Pavilion Ballroom AB
Evolutionary Computation in Bioinformatics

Scope
Bioinformatics and computational biology present a number of difficult optimization problems with large search spaces. Recent applications of evolutionary computation (EC) in this area suggest that they are well-suited to this area of research. This tutorial workshop will present an introduction to problems in Bioinformatics and introduce some basic Biology. Its focus will be on presenting how EC can be used in certain domains (such as RNA folding) to provide better solutions than some of the deterministic approaches. Also, we will present the pitfalls of biological data analysis and how to effectively evaluate evolutionary algorithms in the Bioinformatics domain.

Topics:
1. Introduction to Biology and Bioinformatics and some sample problems (1 hour)
2. Detailed analysis of RNA folding with Evolutionary Algorithms (45 minutes)
3. Workshop discussions: How to get started in Bioinformatics? What problems is the audience interested in? What have they tried? Achievements? Setbacks? How to foster relationships with life science researchers? How to model
representations and operators to meet biological constraints and features. Evaluation methods. (45 minutes)

This tutorial is intended to get EC researchers or practitioners started in bioinformatics.