A novel result on H∞ performance state estimation for Markovian neural networks with time-varying transition rates

• Published in: Neural computing & applications Vol. 33; no. 24; pp. 17001 - 17011
• Main Authors: Tian, Yufeng, Wang, Zhanshan
• Format: Journal Article
• Language: English
• Published: London Springer London 01.12.2021; Springer Nature B.V
• Subjects: Apexes; Artificial Intelligence; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Data Mining and Knowledge Discovery; H infinity; Image Processing and Computer Vision; Linear matrix inequalities; Mathematical analysis; Neural networks; Original Article; Parameters; Probability and Statistics in Computer Science; State estimation; Time-varying transition rates; Delay-dependent activation function inequalities; State estimation; Switched vertices approach; Markovian neural networks
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-021-06291-1

This paper is concerned with the H ∞ performance state estimation for Markovian neural networks (MNNs) with time-varying transition rates. The time-varying TRs are considered to be in a polytopic sense. To fully consider the time-varying transition rates, a parameter-dependent Lyapunov–Krasovskii functional is constructed. A switched vertices approach is proposed to remove the bound assumptions on the derivative of time-varying parameters in the previous works. To fully facilitate the delay information and the slope information of activation function, some delay-dependent activation function inequalities are constructed under a separation of the estimation error of activation function. Based on these ingredients, a novel result on H ∞ performance state estimation for MNNs is presented in terms of linear matrix inequalities, which is more practical and less conservative. An example demonstrates the effectiveness of proposed approaches.