The Parallel Adaptation of Artificial Feed-Forward Neural Networks

Bc. Marek Libra

Master's thesis

Abstract:

The distribution of an artificial feed-forward neural network adaptation in a computer network on low cost x86 PCs interconnected by Internet is proposed. Recently such a distributed environment is well available for both the academic and commercial sphere, so my work can have positive impact on practical applicability in solving real-world problems. The combination of genetic algorithms with the back propagation algorithm is discussed. The comparison of commonly used serial adaptation techniques and the suggested distributed approaches is provided. Positive impact of suggested techniques is shown.

Kľúčové slová

Feed-forward neural network, back propagation, distributed adaptation, genetic algorithm

Jazyk práce: Czech

Datum vytvoření / odevzdání či podání práce: 19. 5. 2006

Obhajoba závěrečné práce

• Obhajoba proběhla 28. 6. 2006
• Vedúci: doc. Ing. Jan Žižka, CSc.

Masaryk University

Faculty of Informatics

Master programme / odbor:
Applied Informatics / Applied Informatics