Stochastic sampled-data control for state estimation of time-varying delayed neural networks

Area of Science:

• Control Systems Engineering
• Artificial Neural Networks
• Nonlinear Systems Analysis

Background:

• State estimation is crucial for neural networks, especially with time-varying delays.
• Existing methods often struggle with the complexities of stochastic sampling and time-varying delays.

Purpose of the Study:

• To develop a novel state estimator for neural networks with time-varying delays.
• To address the challenges posed by stochastic sampled-data in neural network state estimation.

Main Methods:

• Utilizing sampled-data with stochastic sampling for estimator design.
• Implementing a novel approach by dividing activation function bounding into two subintervals.
• Proposing a discontinuous Lyapunov functional based on the extended Wirtinger inequality to leverage sawtooth delay characteristics.

Main Results:

• The proposed state estimator design is formulated using linear matrix inequalities (LMIs).
• The method effectively handles time-varying delays and stochastic sampling in neural networks.
• Numerical examples demonstrate the superior performance of the developed state estimator.

Conclusions:

• The novel state estimation approach provides an effective solution for neural networks with time-varying delays.
• The use of discontinuous Lyapunov functionals and specific bounding techniques enhances estimation performance.
• The LMI-based characterization offers a computationally tractable method for designing the estimator gain.