Optical Realizations of Self-Organizing Neural Networks

  • Cornelia Denz


Among the architectures and algorithms suggested for neural network implementations, the self-organizing map (see section 2.4 and [382]) has the property of creating spatially organized internal representations of various input signals or patterns. It does this by taking the spatial neighbourhood of the cells into account during learning. In its basic version a cell or a group of cells becomes specifically tuned to various input patterns or classes of input patterns through an unsupervised learning process. The spatial location of a cell in the network corresponds to a particular set of input patterns. The spatial clustering of the cells and their organization into topologically related subsets then result in a high degree of classification efficiency. This is particularly relevant in cases in which no a priori information about the scaling of the problem is known, a fact that argues strongly in favour of the use of an optically addressable, continuous adaptive medium for implementation of the neural activity.


Input Pattern Reference Beam Competitive Learning Photorefractive Crystal Neural Firing 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Further Reading

  1. 1.
    R.A. Athale, J. Davis (eds.), Neural network models for optical computing, Proc. SPIE 882 (1988). An older conference proceeding on optical neural network realizations, with several important contributions on self-organized and competitive learning.Google Scholar

Copyright information

© Springer Fachmedien Wiesbaden 1998

Authors and Affiliations

  • Cornelia Denz
    • 1
  1. 1.Institut für Angewandte Optik, AG Photorefraktive OptikTechnische Universität DarmstadtDarmstadtGermany

Personalised recommendations