Synchronization of complex-valued neural networks: A complete sliding mode control approach
Visualitza/Obre
thesis-di-palo.pdf (2,949Mb) (Accés restringit)
Estadístiques de LA Referencia / Recolecta
Inclou dades d'ús des de 2022
Cita com:
hdl:2117/368600
Realitzat a/ambUniversità degli Studi di Napoli Federico II
Tipus de documentTreball de recerca tutelat
Data2022-05-03
Condicions d'accésAccés restringit per decisió de l'autor
Tots els drets reservats. Aquesta obra està protegida pels drets de propietat intel·lectual i
industrial corresponents. Sense perjudici de les exempcions legals existents, queda prohibida la seva
reproducció, distribució, comunicació pública o transformació sense l'autorització del titular dels drets
Abstract
Artificial neural networks (ANNs) are data-driven computing systems inspired by the dynamics and functionality of the human brain. A neural network can be represented as a complex system composed by a high number of interconnected neurons (∼ 1011). A typical dynamical behavior of neural networks is synchronization, which has become an important research issue for its wide applications in many fields, such as secure communication, image encryption, associative memory and pattern recognition. Most results about the dynamic analysis of neural networks have focused on real-valued neural networks. However, complex-valued neural networks (CVNNs) have advantages in different application fields. CVNNs are neural networks where the state space is the n-dimensional set of complex numbers, Cn, and also the other variables involved, such as inputs, outputs and parameters, are represented by complex numbers. In many practical applications complex numbers are often used such as in telecommunications, robotics, image processing, sonar, radar and speech recognition, so CVNNs have potential applications in these domains. One of the most important advantages of CVNNs is good compatibility with wave phenomena. In general, propagation and interference of electromagnetic waves is expressed by the magnitude of transmission and reflection, phase progression and retardation, superposition of fields, etc. These phenomena are expressed simply and naturally by the use of complex numbers. The objective of this work is to develop sliding mode control (SMC) strategies in order to achieve synchronization in CVNNs. The thesis will be structured as follows: In chapter 2 an overview on neural networks is reported with a focus on the complex-valued ones, then the dynamic model of the target network is derived. In chapter 3 we give the definition of synchronization in CVNNs, and the master stability equations approach for complex-valued systems is presented. In chapter 4 two types of sliding mode controllers are developed, with the purpose of achieving synchronization in the network. The first one is a simplified version of a SMC available in the literature [6], designed considering the n-dimensional complex-valued network as a 2n-dimensional real-valued network. The second, innovative, is an extension to the MIMO case of a fully complex-valued SMC presented in, which improves the performances obtained in the previous case. In chapter 5 we carry out the validation of the control strategies developed in the previous one. A specific complex-valued neural network is considered and the controllers are implemented in M atlab and Simulink. In chapter 6 conclusions are drawn
MatèriesSelf-organizing systems -- Software -- Design and construction, Sliding mode control -- Software -- Design and construction, Sistemes autoorganitzatius -- Programari -- Disseny i construcció, Control en mode lliscant -- Programari -- Disseny i construcció
TitulacióMOBILITAT INCOMING
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
thesis-di-palo.pdf | 2,949Mb | Accés restringit |