Bio Inspired Algorithms
- Department of Cybernetics
The students will learn some of the uncoventional methods of computational intelligence aimed at solving complex tasks of classification, modeling, clustering, search and optimization. Bio-inspired algorithms take advantage of analogies to various phenomena in the nature and society. The main topics of the subject are artificial neural networks and evolutionary algorithms.
Basic knowledge of optimization
- Syllabus of lectures:
1.Introduction -- relations to conventional optimization methods, black-box optimization, randomized search methods.
2.Introduction to artificial neural networks, history, typical tasks and their solutions, types of neural networks learning. Perceptron.
3.Supervised learning -- approximation and classification, local and global units in neural networks. Multi-layered perceptron, RBF networks, GMDH networks.
4.Unsupervised learning -- clustering with neural networks, self-organization, Hebb's rule, Hopfield network, associative memory, ART networks.
5.Kohonen's self-organizing map (SOM), competitive learning, reinforcement learning.
6.Error back-propagation algorithm, universal approximation, Kolmogorov theorem.
7.Temporal sequences processing, recurrent neural networks, Elman network, back-propagation through time.
8.Simple genetic algorithm (SGA) -- history, basic cycle, genetic operators, schema theorem.
9.Evolutionary algorithms with real representation -- evolutionary strategy, crossover operators. Differential evolution (DE).
10.Neuroevolution -- evolutionary techniques for neural network structure learning and parameter tunning. NEAT system.
11.Multiobjective optimization -- dominance principle, Pareto-optimal solutions, multiobjective evolutionary algorithms (NSGA-II, SPEA2).
12.Genetic programming (GP) -- tree representation, initialization, operators, strongly-typed GP, automatically defined functions (ADF).
- Syllabus of tutorials:
1.Seminar organization. Black box neural network (MLP), approximation, classification, local search examples in Matlab.
2.Neural network software, Mathematica, Weka.
3.First assignment introduction (introduction to data).
4.Elaboration of the first assignment.
5.Elaboration of the first assignment.
6.First assignment presentation and evaluation.
7.Second assignment introduction (evolutionary algorithms).
8.Simple genetic algorithm (SGA). Influence of SGA parameters on its behaviour. Examples of evolutionary algorithms in Matlab.
9.Elaboration of the second assignment.
10.Successful applications of evolutionary algorithms.
12.Second assignment hand-in and evaluation.
- Study Objective:
The goal of this subject is to acquaint students with unconventional methods of computational intelligence aimed at solving complex tasks of classification, modeling, clustering, search and optimization. The main topics of the subject are artificial neural networks and evolutionary algorithms.
- Study materials:
1.Haykin, S.: Neural Networks: A Comprehensive Foundation, 2nd edition, Prentice Hall, 1998
2.Rojas, R.: Neural Networks: A Systematic Introduction, Springer, 1996
3.Michalewicz, Z.: Genetic Algorithms + Data Structures = Evolution Programs, Springer, 1998
4.Michalewicz, Z.: How to solve it? Modern heuristics. 2nd ed. Springer, 2004.
- Further information:
- No time-table has been prepared for this course
- The course is a part of the following study plans: