Title :
Hopfield networks as discrete dynamical systems
Author :
Desilva, Christopher J S ; Attikiouzel, Yianni
Author_Institution :
Dept. of Electr. & Electron. Eng., Univ. of Western Australia, Nedlands, WA, Australia
Abstract :
A geometrical approach based on dynamical systems theory is presented for the analysis of Hopfield networks. Basic definitions from dynamical systems theory are presented and illustrated with a simple example. The eigenspaces of the weight matrix, the geometry of the energy manifold, and the diffeomorphisms induced by the sphere are presented as tools for the study of network dynamics. Some simple results are given. A number of topics for further research are proposed
Keywords :
Hopfield neural nets; discrete systems; geometry; Hopfield neural network analysis; diffeomorphisms; discrete dynamical systems; eigenspaces; energy manifold; geometrical approach; weight matrix; Geometry; Information analysis; Information processing; Information theory; Intelligent networks; Intelligent systems; Linear algebra; State-space methods; Steady-state;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.227183
