Accepted:
00/00/1994
| |
/vol01/bouzer01/ | © Copyright 1994 | |||
| Volume 01 | Received: Accepted: |
00/00/1994 00/00/1994 |
|||
|
Convergence of symmetric shunting competitive neural networks
Bouzerdoum, A. |
|
| Abstract | |
| A dynamical analysis of a class of recurrent shunting competitive neural networks is presented. These networks are bounded-input bounded-output stable dynamical systems. Furthermore, if the connection matrix is symmetric, they admit a global Liapunov function in the entire n-space R^n. Using LaSalle invariance principle, it is shown that symmetric recurrent shunting competitive neural networks are convergent dynamical systems; that is, all trajectories converge to equilibria. Thus, these networks cannot sustain oscillations; however, they are capable of exhibiting bifurcation phenomena. In particular, a network is presented which exhibits the pitchfork bifurcation as the input pattern is varied. | |
| Full Text |
|
|
Multimedia Links (none) Reference Links (none) Citation Reference Bouzerdoum, A. (1994), Convergence of symmetric shunting competitive neural networks, Complexity International, Volume 01, Paper ID: bouzer01, URL: http://www.complexity.org.au/vol01/bouzer01/ |
Get viewers for PS & PDF   |
