Reduced-order state estimation of delayed recurrent neural networks

He Huang; Tingwen Huang; Xiaoping Chen

doi:10.1016/j.neunet.2017.11.002

Reduced-order state estimation of delayed recurrent neural networks

Neural Netw. 2018 Feb:98:59-64. doi: 10.1016/j.neunet.2017.11.002. Epub 2017 Nov 10.

Authors

He Huang¹, Tingwen Huang², Xiaoping Chen³

Affiliations

¹ School of Electronics and Information Engineering, Soochow University, Suzhou 215006, PR China. Electronic address: hhuang@suda.edu.cn.
² Texas A&M University at Qatar, Doha 5825, Qatar. Electronic address: tingwen.huang@qatar.tamu.edu.
³ School of Electronics and Information Engineering, Soochow University, Suzhou 215006, PR China. Electronic address: xpchen@suda.edu.cn.

PMID: 29179153
DOI: 10.1016/j.neunet.2017.11.002

Abstract

Different from the widely-studied full-order state estimator design, this paper focuses on dealing with the reduced-order state estimation problem for delayed recurrent neural networks. By employing an integral inequality, a delay-dependent design approach is proposed, and global asymptotical stability of the resulting error system is guaranteed. It is shown that the gain matrix of the reduced-order state estimator is determined by the solution of a linear matrix inequality. Numerical examples are provided to illustrate the effectiveness of the developed result.

Keywords: Global asymptotical stability; Recurrent neural networks; Reduced-order state estimation; Time delay.

MeSH terms

Neural Networks, Computer*
Time Factors