Neural networks with dynamical threshold
For a neural network with sign-constrained weights weights three types of attractor can affect the dynamics: retrieval, spurious and uniform attractors. The uniform attractors can dominate the dynamics if there is a substancial weight-sign bias. We will show that it is possible to define dynamical thresholds for various learning rules which can eliminate uniform attracting states for any value of the weight-sign bias.
Year of publication: |
1992
|
---|---|
Authors: | Campbell, C. ; Wong, K.Y.M. |
Published in: |
Physica A: Statistical Mechanics and its Applications. - Elsevier, ISSN 0378-4371. - Vol. 185.1992, 1, p. 378-384
|
Publisher: |
Elsevier |
Saved in:
Saved in favorites
Similar items by person
-
Statistical mechanics of neural networks: what are the differences between wide and narrow basins?
Wong, K.Y.M., (1992)
-
Freezing transitions in neural networks
Wong, K.Y.M., (1994)
-
Probing the attractors in neural networks
Wong, K.Y.M., (1993)
- More ...