Stability analysis of autonomous ratio-memory cellular nonlinear networks for pattern recognition

Authors:
Su-Yung Tsai;Chi-Hsu Wang;Chung-Yu Wu
Affiliations:
Nanoelectronics and Gigascale Systems Laboratory, Department of Electronics Engineering & Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan;Department of Electrical Engineering, National Chiao Tung University, Hsinchu, Taiwan;Nanoelectronics and Gigascale Systems Laboratory, Department of Electronics Engineering & Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan
Venue:
IEEE Transactions on Circuits and Systems Part I: Regular Papers
Year:
2010

Citing 8
Cited 0

Introduction to the theory of neural computation

Introduction to the theory of neural computation
Weakly connected neural networks

Weakly connected neural networks
A synthesis procedure for associative memories based on space-varying cellular neural networks

Neural Networks
Learning on Silicon: Adaptive VLSI Neural Systems

Learning on Silicon: Adaptive VLSI Neural Systems
Spiking Neuron Models: An Introduction

Spiking Neuron Models: An Introduction
Analog VLSI: Circuits and Principles

Analog VLSI: Circuits and Principles
Eigenvector methods for automated detection of electrocardiographic changes in partial epileptic patients

IEEE Transactions on Information Technology in Biomedicine - Special section on biomedical informatics
CMOS current-mode neural associative memory design with on-chip learning

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

The stability analysis via the Lyapunov theorem for Autonomous Ratio-Memory Cellular Nonlinear Networks (ARM-CNNs) is proposed. A conservative domain of attraction (DOA) is found from the stability analysis through a graphical method without complicated numerical analysis. The stability analysis shows that ARMCNNs can tolerate large ratio weight variations. This paper also presents the ARMCNN with self-feedback (SARMCNN) to overcome the problem of isolated neurons due to low correlation between neighboring neurons. The SARMCNN recognition rate (RR) is compared with other CNN constructed via the singular value decomposition technique (SVD-CNN).