TOP

Soft Computing

Autumn 2005

Monday and Wednesday 16:30 ~ 17:45

Electronic & Information Building  #107

last updated: 2005/12/22

1. Course Description

This course will provide students the basic concepts of different methods and tools for processing of uncertainty in intelligent systems, such as, fuzzy models, neural networks, probabilistic models, and foundations of its using in real systems. The course cover main concepts of philosophy of AI and hybrid intelligent systems, classification and architecture of hybrid intelligent systems.

 

2. Textbooks

* Presentation materials   [HERE]

* Contents of lectures:

1. Introduction

             What is Soft Computing (SC)?

             Necessary of Representation of Fuzzy Things in AI

Two approaches to development of AI – Neuroinformatics and Logical AI

2. Two-level model of mind

             Two levels – associative (images) and logical (signs)

             Model of associative thinking

             Analogy and associations

             Tasks of recognition, classification and clusterization of images

             Two approaches of improvement of logical and neural models – “From logic to fuzziness” (up-down) and “From neural networks to concepts” (down-up)

3. From logic to fuzziness

Fuzzy Sets

Fuzzy Logic

Linguistic Variables

Pseudo-Physics logics

                           Definitions

                           Space Pseudo-Physics logic

                           Temporal Pseudo-Physics logic

             Problem of learning of intelligent systems based on fuzzy sets

4. From neural networks to concepts

             Main definitions

             Classification of Neural Networks

             Kinds of tasks for solving by NN

             Kinds of Training of NN

             Perceptron and Error Back Propagation Algorithm

             Hopfild’s network and Hebb’s training algorithm

             Multi-Layer recurrent neural networks

             SOM of Kohonen

             Reinforcement learning

             Constructive algorithms of training of neural networks

                           Hamming’s neural network

                           ART of Grossberg-Carpenter

                           Constructive algorithms of training for MLP

5. Probabilistic methods of Knowledge Representation and Reasoning

             Bayes’s nets

             Genetic Algorithms

6. Using of neural networks in NLP

6. Paradigm of Hybrid Intelligent Systems (HIS)

             Motivation

             Kinds of architectures

             Fuzzy neural networks

             Review of HIS

7. Introduction to generalized theory of uncertainty

8. Conclusion. Modern trends in development of soft computing

 

3. Grading Police (tentative)

Midterm

Final

Total

50%

50%

100%

 

4. Instructor

Name

Andrey Gavrilov  (Russia)

Contact Information

Tehephone : 031-201-2493
E-mail : avg@oslab.khu.ac.kr
Office :

Electronic & Information Building #B08

_material     

5. Tentative Schedule (updating)

#

Kind

Topic

Material

1

Lecture 1

Introduction. What is Soft Computing (SC)?

Necessary of Representation of Fuzzy Things in AI. Two approaches to development of AI – Neuroinformatics and Logical AI.

DOWN

2

Lecture 2

Two-level model of mind.    Two levels – associative (images) and logical (signs).

Model of associative thinking. Analogy and associations.    Tasks of recognition, classification and clustering of images. Two approaches of improvement of logical and neural models – “From logic to fuzziness” (top-down) and “From neural networks to concepts” (bottom-up).

DOWN

3

Lecture 3

Fuzzy Sets. Fuzzy Logic

DOWN

4

Colloquium 1

Problems of simulation of mind.

DOWN

5

Lecture 4

Linguistic Variables. Pseudo-Physics logics Definitions. Space Pseudo-Physics logic. Temporal Pseudo-Physics logic.

DOWN

6

Lecture 5

Using of Fuzzy logic in control systems.

DOWN

7

Lecture 6

Introduction in neural networks. First models of NN.

Tasks solved by NN. Kinds of learning of NN.

DOWN

8

Lecture 7

Multilayer perceptrons (MLP). Tasks of regression and classifications. Algorithm error back propagation.

DOWN

9

Lecture 8

Using of perceptron for image recognition and forecasting.

DOWN

10

Lecture 9

Associative memory based on Hopfield model

DOWN

11

Lecture 10

Boltzmann Machine

DOWN

12

Lecture 11

Multi-layer recurrent neural networks.

DOWN

13

Lecture 12

Self-organized map (SOM) of Kohonen

DOWN

14

Lecture 13

Reinforcement learning and it’s using in robotics.

DOWN

15

Lecture 14

Constructive algorithms of learning of NN.

DOWN

16

Lecture 15

Hemming’s Newral Networks.Constructive algorithms of learning for MLP.

DOWN

17

Lecture 16

Introduction to spike neuron models. Tools for developments of neural networks

DOWN

18

Colloquium 2

How to select of neural model for solving of task? Hybrid neural networks.

DOWN

19

Exam

MIDTEAM EXAM

PREPARE

20

Lecture 17

Introduction to probabilistic reasoning. Bayes’s nets. Markov’s model.

DOWN

21

Lecture 18

Foundations of genetic algorithms (GA). Using of GA.

DOWN

22

Lecture 19

Motivation and main definitions of Hybrid intelligent systems (HIS). Kinds of architectures. Fuzzy neural networks. Two-hemisphere architecture.

PART_I

PART_II

23

Lecture 20

Review of HIS.

DOWN

24

Colloquium 3

How select of hybrid architecture for solving of task?

 

25

Lecture 21

Review of using of NN for solving of real tasks.

DOWN

26

Lecture 22

Using of NN in NLP and speech recognition

DOWN

27

Lecture 23

Hardware neural networks

DOWN

28

Lecture 24

Future of soft computing. Introduction to generalized theory of uncertainty

DOWN

29

Exam

FINAL EXAM

 

 

6. Suggested reading

1.  Arbib, M. A. The Metaphorical Brain. New York: Wiley (1990).

2.  Computationally Intelligent Hybrid Systems: The Fusion of Soft Computing and Hard Computing. Seppo J. Ovaska (Editor). (Electronic contents and Chapter 1 are available).

3.  Duda R.O., Hart P.E., Stock D.G. Pattern Classification, John Wiley (2001).

4.  Honavar, V. & Uhr, L. (Ed.) Artificial Intelligence and Neural Networks: Steps Toward Principled Integration. San Diego, CA: Academic Press (1994).

5.  Hristev R.M. Artificial Neural Networks. (Electronic book is available).

6.  Jones M.T. AI application Programming. Charles River Media, Inc., Hingham, Massachusetts (2003).

7.  Kecman V. Learning and Soft Computing. MIT, 2001. (Electronic book is available).

8.  Koivo H. Soft Computing in dynamical systems. (Electronic book is available).

9.  Krose B., van der Smagt P. An Introduction to Neural Networks. 1996 (Electronic book is available).

10. Lawrence J. Introduction to Neural Networks. California Scientific, Nevada City, (1993).

11. Luger G.F. Artificial Intelligence. Structures and Strategies for Complex Problem Solving. Addison Wesley  (2002).

12. Mitchell T.M. Machine learning. McGraw Hill, 1997.

13. Nillson N.J. Introduction to Machine Learning. Draft of textbook. (Electronic book is available).

14. Russel, S. & Norvig, P., Artificial Intelligence - a modern approach (2th edition). Englewood Cliffs, NJ: Prentice Hall (2002). (Electronic contents and ?hapters 5,7,11,20 are available).

15. Sutton R.S., Barto A.J. Reinforcement Learning: An Introduction. The MIT Press Cambridge, Massachusetts, London. (Electronic book is available).

16. Wasserman, P. Advanced Methods in Neural Computing. New York: Van Nostrand Rheinhold (1993).

17. Wilson R.A. Quantum Psychology. 1990.

 

7. Download the syllabus file (MS-WORD) :      [DOWN]

 

topicon.gif