Neural network synthesis via asynchronous analytic programming

  • Authors:
  • Pavel Vařacha

  • Affiliations:
  • Tomas Bata University in Zlin, Faculty of Applied Informatics, Zlin, Czech Republic

  • Venue:
  • NNECFSIC'12 Proceedings of the 12th WSEAS international conference on Neural networks, fuzzy systems, evolutionary computing & automation
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

This article deals with Analytic Programming (AP) which was proven to be highly effective tool of Artificial Neural Network (ANN) synthesis and optimization. New innovative asynchronous distribution of Self-Organizing Migration Algorithm (SOMA) is introduced and used together with AP. Such implementation can for example save 67% of computation time if distributed between 8 processors. Efficiency of AP as well as asynchronous distribution of SOMA was tested and statistically measured on 921937 evaluations, each of them containing another separate execution of SOMA.