Journal article
Stability of a class of directionally sensitive asymmetric nonlinear neural networks
Neural networks, v 9(4), pp 555-565
1996
Featured in Collection : UN Sustainable Development Goals @ Drexel
Abstract
Asymmetrically connected inhibitory shunting networks are believed to occur in many areas of the brain. One such area is the visual system; these networks are found useful in explaining many peripheral visual phenomena. This paper examines a class of asymmetrically connected networks for stability and uniqueness of outputs; both time-invariant and moving inputs are considered.
Under different sets of constaints upon the network characteristics and operating region, it is possible to obtain:
•
• bounded-input-bounded-output stability of the solutions;
•
• stability and uniqueness of the output to any input for a network of finite extent, with time-invariant inputs giving time-invariant outputs;
•
• localization of edge effects on a finite network, and stability and uniqueness of the output to any input for a network of infinite extent, with time-invariant or traveling-wave inputs giving time-invariant or traveling-wave outputs respectively.
The constraints required for these three sets of characteristics are successively more stringent. The most interesting constraints are those required for localization of edge effects on a finite network; they are not satisfied by a network in which the output from each node is obtained by sharply thresholding the node potential. However, they can be satisfied if the network operates in a relatively low signal region, below saturation of the outputs. An example of a network which does satisfy the constraints is given by certain field effect transistor (FET) implementations.
Metrics
Details
- Title
- Stability of a class of directionally sensitive asymmetric nonlinear neural networks
- Creators
- S.P. Tonkin - University of WashingtonR.B. Pinter - University of WashingtonB. Nabet - Drexel University
- Publication Details
- Neural networks, v 9(4), pp 555-565
- Publisher
- Elsevier
- Resource Type
- Journal article
- Language
- English
- Academic Unit
- Electrical and Computer Engineering
- Web of Science ID
- WOS:A1996UT30900001
- Scopus ID
- 2-s2.0-0030176269
- Other Identifier
- 991019169560704721
UN Sustainable Development Goals (SDGs)
This publication has contributed to the advancement of the following goals:
InCites Highlights
Data related to this publication, from InCites Benchmarking & Analytics tool:
- Collaboration types
- Domestic collaboration
- Web of Science research areas
- Computer Science, Artificial Intelligence
- Neurosciences