Basins of attraction of neural network models
Abstract
The basins of attraction of memory states of a neural network model are explored by taking random slices through the state space. As the number of stored states increases, the basins of attraction become very complicated. This phenomena substantially reduces the performance of the network as a content addressable memory. It is shown that the performance of the network can be improved by algorithms which smooth the basins of attraction.
- Publication:
-
Neural Networks for Computing
- Pub Date:
- August 1986
- DOI:
- 10.1063/1.36277
- Bibcode:
- 1986AIPC..151..259K
- Keywords:
-
- 89.80.+h;
- 84.20.Pc