Global Stabilization of Memristive Neural Networks with Leakage and Time-Varying Delays Via Quantized Sliding-Mode Contr
- PDF / 716,868 Bytes
- 18 Pages / 439.37 x 666.142 pts Page_size
- 101 Downloads / 198 Views
Global Stabilization of Memristive Neural Networks with Leakage and Time-Varying Delays Via Quantized Sliding-Mode Controller Yuting Cao1 · Bo Sun2 · Zhenyuan Guo1 · Tingwen Huang3 · Zheng Yan4 · Shiping Wen4 Accepted: 17 September 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract This pape investigates the global stabilization of memristive neural networks (MNNs) with leakage and time-varying delays via quantized sliding-mode controller. The leakage delay is considered in the MNNs. Sliding mode controller is imported to ensure global stabilization of delayed MNNs. We also introduce two quantization schemes with uniform quantizer and logarithmic quantizer. Our goal is to deal with errors before and after quantization. We give some simulations and comparisons between two quantizers in the end of this paper. Keywords Memristive neural network · Time-varying delay · Leakage delay · Global stabilization · Quantized sliding-mode controller · Quantization scheme
1 Introduction An artificial neural network is an information processing system designed to mimic the structure of the human brain and its functions [1–7]. The intensity of the synaptic connections between neurons is variable, which is the basis of learning and memory. Artificial neural networks can be self-learning and adaptive through “training” [8,9]. The key to neural network technology is weight design. The hardware implementation of weights requires a nanoscale component that maintains memory for a long time and does not consume energy. The traditional artificial neural network technology is based on the traditional computer. The main disadvantage is that the computation is huge and the operation is not parallel processing. If the parallel distributed processing, nonlinear processing, self-learning function and adaptive
B
Shiping Wen [email protected]
1
College of Mathematics and Econometrics, Hunan University, Changsha 410082, China
2
School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
3
Science Program, Texas A&M University at Qatar, Doha 23874, Qatar
4
Australian AI Institute, University of Technology, Sydney, NSW, Australia
123
Y. Cao et al.
function of the artificial neural network are realized on the hardware, the shortcomings of the artificial neural network on the traditional computer can be solved. A single memristor can simulate the synaptic function, and the memristor can be easily combined with nanocross-connect technology, with the advantages of large-scale parallel processing, distributed information storage, and huge storage capacity. Therefore, the use of the memristive system is the best way to simulate the synaptic function of artificial neural networks, and thus has become a hot topic in recent years [10–12]. As early as 1971, Chua [13] predicted through theoretical calculations that there must be a fourth passive electronic component, namely a memristor, in addition to resistance, capacitance, and inductance. The
Data Loading...
