Kanpur Genetic Algorithms Laboratory

A week-long course entitled
Evolutionary Algorithms for Optimization
at Helsinki School of Economics, Finland
23-27 May 2005

Contents:

Course Instructor
Course Dates
Course Description
Course Topics
Prerequisites
Course Handouts
Evaluation
EA Implementation
Further Reading Materials

Course Instructor

Take advantage of learning and understanding the fast-growing field of evolutionary computation from someone who has almost 18 years of research and teaching experience in the field and has written two popular text books and over 150 international journal and conference papers:

Prof. Kalyanmoy Deb
Department of Mechanical Engineering
Indian Institute of Technology Kanpur
Kanpur, PIN 208016, India
Email: deb@iitk.ac.in
http://www.iitk.ac.in/kangal

Prof. Deb is a fellow of Intl. Society of Genetic and Evolutionary Computation (ISGEC) and Indian National Academy of Engineering (INAE). He is also the associate editor of two leading journals in the evolutionary computation field: IEEE Trans. on Evolutionary Computation and Evolutionary Computation Journal from MIT Press and in the editorial board of a few other journals including Genetic Programming and Evolvable Machines and Engineering Optimization.

Course Dates

23-27 May, 2005

Course Description

Over the past two decades, evolutionary algorithms (EAs) - search and optimization algorithms developed based on working principles of natural evolution - have emerged as alternate optimization algorithms for solving different kinds of optimization problems often encountered in sciences, engineering and commerce. In this week-long course, a gentile introduction to evolutionary algorithm, its working principles, mathematical underpinnings, its extensions to solve different kinds of optimization problems including nonlinear constraint handling, multi-objective optimization and scheduling problems will be introduced. Some real-world case studies will also be presented to illustrate the practical importance of EAs.

At the end of the course, the participants are likely to know the fundamentals of evolutionary algorithms and their application domain and a niche of problems where EAs are most suitable. Besides, they will have relevant important research papers/documents and computer codes so that they will be in a position to start reading EA papers and begin working in the fast-growing area of evolutionary computation.

Course Topics

About 10 one-and-half-hour lectures (a total of about 15 hours of core teaching) is planned as follows:

Total Lecture Hours: 17 Hours (23-27.05.2005)

(23 May 2005 9:00-10:30 Hrs) Motivation for evolutionary algorithms (EAs) and a gentle introduction to EAs

1.1 Difficulties with common classical methods

1.2 Nature and optimization

1.3 EA constituents and early EA studies

1.4 A vanilla genetic algorithm with binary coding

1.5 An application on a test problem

1.6 An application case study -- car suspension design

(23 May 2005 11:00-12:30 Hrs) Handling constraints (linear and nonlinear) with EAs

2.1 Five different procedures

2.2 A penalty-parameter-less approach

2.3 Simulation results on test problems

(24 May 2005 9:00-10:30 Hrs) Knowledge-assisted evolutionary algorithms and some theory of EAs

3.1 No free lunch (NFL) theorem and the need of knowledge-assisted EAs

3.2 Knowledge-assisted initialization and EA operators

3.3 A casting scheduling problem and advantages of a knowledge-assisted EA

3.4 Theoretical underpinnings of GAs

(24 May 2005 11:00-12:30 Hrs) Real-parameter optimization using EAs

4.1 Difficulties with binary-coded GAs or EAs

4.2 Real-parameter GAs

4.3 Simulated binary crossover (SBX) and other recombination and mutation operators

4.4 Simulation results on test problems

4.5 Self-adaptive GAs

(24 May 2005 15:00-16:00 Hrs) Simple GA implementational issues

(25 May 2005 9:00-10:30 Hrs) Evolution Strategies (ESs)

5.1 (1+1)-ES and early applications

5.2 (mu+,lambda)-ES

5.3 Recombinative ES

5.4 Self-adaptive ESs

(25 May 2005 11:00-12:30 Hrs) Differential evolution (DE) and Particle Swarm Optimization (PSO)

6.1 Differential evolution

6.2 Particle swarm optimization

6.3 Generalized generation gap (G3) algorithm and parent-centric recombination (PCX)

6.4 Comparison of G3+PCX with different ESs, DE and quasi-Newton method

(26 May 2005 9:00-10:30 Hrs) Multi-modal optimization

7.1 Finding multiple optimal solutions

7.2 Niching methods (Sharing, clearing, clustering etc.)

7.3 Speciation methods (tag-template, mating restriction etc.)

(26 May 2005 11:00-12:30 Hrs) Multi-objective optimization

8.1 Introduction

8.2 Major classical methods

8.3 Philosophy of evolutionary methods in multi-objective optimization

8.4 Domination and niching

8.5 Early methodologies (non-elitist methods) -- NSGA, NPGA, MOGA etc.

8.6 State-of-the-art methodologies (elitist methods) -- NSGA-II, SPEA2, PESA etc.

(26 May 2005 15:00-16:00 Hrs) Multi-objective GA implementational issues

(27 May 2005 9:00-10:30 Hrs) Advanced topics in evolutionary multi-objective optimization (EMO)

9.1 Constraint handling

9.2 Applications -- better decision making

9.3 Finding common principles among Pareto-optimal solutions

9.4 Assisting in other optimization problems

9.5 Interactive EMO

(27 May 2005 11:00-12:30 Hrs) Other optimization problems

10.1 Scheduling EAs -- TSPs, job-shop scheduling, class-room time-tabling etc.

10.2 Goal programming using EAs

10.3 Genetic programming (GP)

10.4 Conclusions and main research areas in EAs

Home Assignment

11.1 Click here to download the home assignment 1 and 2.

In addition to 3-hour lectures each day, on a couple of days (Day 2 and 4) will have 1-hour lecture on EA implementation issues. Thus, there will be a total of 17-hour lectures. Home assignments (hand calculation and computer simulation) will be given at the end of each lecture, which are to be worked on and submitted in the next day.

Prerequisites

A preliminary background on optimization and knowledge on a programming language (preferably C or C++) are anticipated.

Course Handouts

A copy of most lecture materials (PPT) will be given.
Some board teaching will also be done, for which relevant and salient research papers will be given.
Some freely available EA codes will be given. Information about commercial codes, major EA journals and major EA conferences will be provided.
A set of home assignments.

Evaluation

Home assignments will be graded and commented.

There will be a Final Examination at the end of the course.

Implementation of GA and NSGA-II

Further Reading Materials

This page is maintained by Dr. Kalyanmoy Deb. Please contact at deb@iitk.ac.in for any comments.
Page last updated on April 15, 2005 by Santosh Tiwari (tiwaris@iitk.ac.in)