Abstract
Although the gradient method with error back-propagation has proved to be successful at teaching multilayered neural networks to perform many tasks, it has a number of rather unrealistic aspects, especially concerning the comparison with biological nerve nets. It is particularly troublesome in this respect: that complete knowledge of the deviation of the output from the desired reaction is required to determine the adjustment even of neurons in hidden layers far separated from the output layer. It is hard to believe that such extended back-coupling mechanisms can operate in complex biological neural networks.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
The concept of selfish or “hedonistic” neurons was apparently first introduced by A.H. Klopf [K182].
This is a well-known educational principle: in most cases reward of successful behavior makes a much better teacher than punishment of failures. Nevertheless, punishment can serve its educational purpose, if applied sparingly and wisely.
A variant of the Kohonen algorithm has also been used to solve the traveling salesman problem [Fo88c]: an arrangement of neurons originally representing a closed circle is continuously deformed under the pull of the cities to be visited.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1995 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Müller, B., Reinhardt, J., Strickland, M.T. (1995). Unsupervised Learning. In: Neural Networks. Physics of Neural Networks. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-57760-4_15
Download citation
DOI: https://doi.org/10.1007/978-3-642-57760-4_15
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-60207-1
Online ISBN: 978-3-642-57760-4
eBook Packages: Springer Book Archive