Abstract
It has been suggested that long-range lateral connections in the cortex play a contextual role in that they modulate the gain of the response to primary receptive field input. In the first part of this paper I show that a network with a set of such pre-wired connections has a shortterm dynamics that enhances and stabilizes coherent information defined across multiple, non-overlapping receptive fields. In the second part, I suggest a simple Hebbian rule that can develop the required pattern of synaptic strengths and describe two simulations where the networks discover information that is defined only by its coherence across receptive fields.
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
References
S. Becker and G. E. Hinton. Self-organizing neural network that discovers surfaces in random-dot stereograms. Nature, vol. 355:pp. 161–163, 1992.
D. Floreano, J. Kay, and W. A. Phillips. A computational theory of learning visual features via contextual guidance. European Conference on Visual Perception, August, Tübingen, 1995.
J. A. Hirsch and C. D. Gilbert. Synaptic physiology of horizontal connections in cat's visual cortex. Journal of Neuroscience, vol. I:pp. 1800–1809, 1991.
E. Oja. A simplified neuron model as a principal component analyzer. Journal of Mathematical Biology, vol. 15:pp. 267–273, 1982.
W. A. Phillips, J. Kay, and D. M. Smyth. The discovery of structure by multistream networks of local processors with contextual guidance. Network, vol. 6:pp. 225–246, 1995.
W. A. Phillips and W. Singer. In search of common foundations for cortical computation. Submitted to Brain and Behavioural Science, 1995.
J. Schmidhuber and D. Prelinger. Discovering predictable classifications. Neural Computation, vol. 4:pp. 625–635, 1993.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1996 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Floreano, D. (1996). Extraction of coherent information from non-overlapping receptive fields. In: von der Malsburg, C., von Seelen, W., Vorbrüggen, J.C., Sendhoff, B. (eds) Artificial Neural Networks — ICANN 96. ICANN 1996. Lecture Notes in Computer Science, vol 1112. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-61510-5_68
Download citation
DOI: https://doi.org/10.1007/3-540-61510-5_68
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-61510-1
Online ISBN: 978-3-540-68684-2
eBook Packages: Springer Book Archive