Executive Summary : | An artificial neural network (NN) is a synthetic computing model resembling the features and structure of biological NNs. Due to the recent innovations of digital technologies, it is necessary to construct discrete-time NNs (DNNs), which are an analogue model of continuous cases, where the discretization may not conserve the framework of the continuous-time counterpart. Meanwhile, the study on the behaviour of a system in terms of energy it can store or dissipate has a significant research interest since it gives a physical and intuitive interpretation of problems such as system stability. The dissipativity theory, an energy-based property of a dynamical system, provides flexible instruments for system analysis. More specifically, the concept of dissipativity theory has been introduced in control problems for a long time to investigate the robust stability properties of systems subjected to nonlinear or time-varying components. For this motivation, this project aims to analyze the stabilization and extended dissipativity performance for a class of delayed DNNs as in the following process: i) To design a suitable controller that assists in stabilizing the delayed DNNs and updating weight components by accelerating towards convergence. ii) Based on the Lyapunov-Krasovskii functional algorithm, extended dissipativity criteria can be established in the form of linear matrix inequalities that can be solved conveniently by using optimization tools available in the MATLAB environment. By utilizing the dynamic nature of prescribed DNNs, the stabilization and robust performances of various real-life engineering problems such as the Quadruple-Tank Process, pendulum, and wind turbine model will be analyzed. Moreover, in the present digitalized world, discrete-time memristor-based neural networks (DMNNs) (made of bio-inspired neuron oscillatory circuits with nanoscale memristors) have attracted extensive research interest in modelling and neurobiological research. In the last few years, the DMNN has been used widely in digital memory, logic circuits, biological and neuromorphic systems, and it is necessary to analyze the qualitative behaviour of DMNNs. Therefore, using the dissipativity nature of prescribed DMNN, this project will analyze the various types of encryption algorithms, such as position permutation related algorithm, value transformation-related algorithm and position-substitution based algorithm. Moreover, this project will aim to develop a new algorithm to transform the desired images more securely. |