Course detail

Evolution Algorithms

FEKT-MEALAcad. year: 2013/2014

The course is focused on deterministic and stochastic optimization methods for finding global minima. It focuses on evolutionary algorithms with populations such as genetic algorithms, controlled random search, evolutionary strategies, particle swarm method, the method of ant colonies and more.

Language of instruction

Czech

Number of ECTS credits

5

Mode of study

Not applicable.

Guarantor

doc. Ing. Jiří Kozumplík, CSc.

Department

Department of Biomedical Engineering (UBMI)

Learning outcomes of the course unit

The graduate of the course is capable of:
Implement a simple analytical optimization method (steepest descent and Newton's method)
To implement the simplex method for finding global extreme
Explain the nature of stochastic optimization methods with populations
Explain the nature of binary and continuous genetic algorithms and the basic operations

Prerequisites

The knowledge on the Bachelor´s degree level is requested, namely on numerical mathematics. The laboratory work is expected knowledge of Matlab programming environment.

Co-requisites

Not applicable.

Planned learning activities and teaching methods

Teaching methods include lectures and computer laboratories. Course is taking advantage of e-learning system. Students have to write a single project/assignment during the course.

Assesment methods and criteria linked to learning outcomes

Requirements for completion of a course are elaborated by the lecturer responsible for the course every year;
- 30 points can be obtained for activity in the laboratory exercises, consisting in solving tasks (for the procedure for the examination must be obtained at least 15 points)
- 70 points can be obtained for the written exam (the written examination is necessary to obtain at least 35 points)

Course curriculum

1. Optimization based on mathematical analysis, optimality conditions, gradient, Hessian
2. Method of steepest descent, Newton's method
3. Stochastic algorithms for finding global minima, the simplex method
4. Evolutionary algorithms with populations. Binary genetic algorithms.
5. Continuous genetic algorithms.
6. Controlled random search, evolutionary strategies, particle swarm
7. Differential evolution, SOMA, ant colony
8. contestants heuristics
9. Test function for checking optimization algorithms
10. Simple methods: climbing algorithm, prohibited search, simulated annealing
11. Experimental comparison of evolutionary algorithms
12. Introduction to genetic programming

Work placements

Not applicable.

Aims

Obtaining an understanding about deterministic and stochastic optimization methods. Introduction to the evolutionary algorithms with populations for finding the global extremes multidimensional functions. Introduction to the genetic programming.

Specification of controlled education, way of implementation and compensation for absences

Delimitation of controlled teaching and its procedures are specified by a regulation issued by the lecturer responsible for the course and updated for every year (see Rozvrhové jednotky).
Basically:
- obligatory computer-lab tutorial (missed labs must be properly excused and can be replaced after agreement with the teacher)
- voluntary lecture.

Recommended optional programme components

Not applicable.

Prerequisites and corequisites

Not applicable.

Basic literature

Tvrdík, J.: Evoluční algoritmy. Skripta, Přírodovědecká fakulta Ostravské univerzity, 2004 (CS)

Recommended reading

Hynek, J.: Genetické algoritmy a genetické programování. Grada Publishing, 2008 (CS)

Classification of course in study plans

  • Programme EEKR-M Master's

    branch M-BEI , 2 year of study, winter semester, elective specialised

  • Programme EEKR-M Master's

    branch M-BEI , 2 year of study, winter semester, elective specialised

  • Programme EEKR-CZV lifelong learning

    branch EE-FLE , 1 year of study, winter semester, elective specialised

Type of course unit

 

Lecture

26 hod., optionally

Teacher / Lecturer

doc. Ing. Jiří Kozumplík, CSc.

Syllabus

1. Optimization based on mathematical analysis, optimality conditions, gradient, Hessian
2. Method of steepest descent, Newton's method
3. Stochastic algorithms for finding global minima, the simplex method
4. Evolutionary algorithms with populations. Binary genetic algorithms.
5. Continuous genetic algorithms.
6. Controlled random search, evolutionary strategies, particle swarm
7. Differential evolution, SOMA, ant colony
8. contestants heuristics
9. Test function for checking optimization algorithms
10. Simple methods: climbing algorithm, prohibited search, simulated annealing
11. Experimental comparison of evolutionary algorithms
12. Introduction to genetic programming

Exercise in computer lab

13 hod., compulsory

Teacher / Lecturer

Ing. Martin Mézl, Ph.D.

Syllabus

1. Analytical method for finding the local minima.
2. Newton method
3. Nelder-Mead method (simplex method).
4. Binary genetic algorithm (GA). Basic selection, crossover and mutation techniques.
5. GA for finding minima in 2D case. Two point crossover method, roulette-wheel selection.
6. Continuous GA. Differences between binary and continuous GA.
7. Application of GA for rigid image registration.
8. Travel salesman problem (TSP). Basic heuristic algorithms - nearest neighboor algorithm, Greedy algorithm.
9. Permutation GA for solving the TSP.
10. Introduction to Particle swarm optimization (PSO) methods. Ant colony optimization for solving the TSP.
11. Student project - implementation of 0-1 knapsack problem or Packing problem.
12. PSO for finding minima of 2D function.