Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning E-Book

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: IEEE/OUP Series on Electromagnetic Wave Theory (formerly IEEE only), Series Editor: Donald G. Dudley.
• Sprache: Englisch

Seitenzahl: 1168

Veröffentlichungsjahr: 2023

Beschreibung

Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning Authoritative reference on the state of the art in the field with additional coverage of important foundational concepts Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning presents cutting-edge research advances in the rapidly growing areas in optical and RF electromagnetic device modeling, simulation, and inverse-design. The text provides a comprehensive treatment of the field on subjects ranging from fundamental theoretical principles and new technological developments to state-of-the-art device design, as well as examples encompassing a wide range of related sub-areas. The content of the book covers all-dielectric and metallodielectric optical metasurface deep learning-accelerated inverse-design, deep neural networks for inverse scattering, applications of deep learning for advanced antenna design, and other related topics. To aid in reader comprehension, each chapter contains 10-15 illustrations, including prototype photos, line graphs, and electric field plots. Contributed to by leading research groups in the field, sample topics covered in Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning include: * Optical and photonic design, including generative machine learning for photonic design and inverse design of electromagnetic systems * RF and antenna design, including artificial neural networks for parametric electromagnetic modeling and optimization and analysis of uniform and non-uniform antenna arrays * Inverse scattering, target classification, and other applications, including deep learning for high contrast inverse scattering of electrically large structures Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning is a must-have resource on the topic for university faculty, graduate students, and engineers within the fields of electromagnetics, wireless communications, antenna/RF design, and photonics, as well as researchers at large defense contractors and government laboratories.

Leseprobe

IEEE Press

445 Hoes Lane Piscataway, NJ 08854

IEEE Press Editorial Board

Sarah Spurgeon,

Editor in Chief

Jón Atli Benediktsson

Behzad Razavi

Jeffrey Reed

Anjan Bose

Jim Lyke

Diomidis Spinellis

James Duncan

Hai Li

Adam Drobot

Amin Moeness

Brian Johnson

Tom Robertazzi

Desineni Subbaram Naidu

Ahmet Murat Tekalp

Advances in Electromagnetics Empowered by Artificial Intelligence and Deep Learning

Edited bySawyer D. Campbell and Douglas H. WernerDepartment of Electrical EngineeringThe Pennsylvania State UniversityUniversity Park, Pennsylvania, USA

Copyright © 2023 by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey.

Published simultaneously in Canada.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, [phone], fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, [phone], fax [phone], or online at http://www.wiley.com/go/permission.

Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at [phone], outside the United States at [phone] or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data Applied for:

Hardback ISBN: 9781119853893

Cover Image and Design: Wiley

To the memory of my mother Joyce L. Campbell

—Sawyer D. Campbell

To my devoted wife Pingjuan Li Werner and to the memory of my grandmother Flora L. Werner

—Douglas H. Werner

About the Editors

Sawyer D. Campbell is an Associate Research Professor in Electrical Engineering and associate director of the Computational Electromagnetics and Antennas Research Laboratory (CEARL), as well as a faculty member of the Materials Research Institute (MRI), at The Pennsylvania State University. He has published over 150 technical papers and proceedings articles and is the author of two books and five book chapters. He is a Senior Member of the Institute of Electrical and Electronics Engineers (IEEE), OPTICA, and SPIE and Life Member of the Applied Computational Electromagnetics Society (ACES). He is the past Chair and current Vice Chair/Treasurer of the IEEE Central Pennsylvania Section.

Douglas H. Werner holds the John L. and Genevieve H. McCain Chair Professorship in Electrical Engineering and is the director of the Computational Electromagnetics and Antennas Research Laboratory (CEARL), as well as a faculty member of the Materials Research Institute (MRI), at The Pennsylvania State University. Prof. Werner has received numerous awards and recognitions for his work in the areas of electromagnetics and optics. He holds 20 patents, has published over 1000 technical papers and proceedings articles, and is the author of 7 books and 35 book chapters. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE), the Institute of Engineering and Technology (IET), Optica, the International Society for Optics and Photonics (SPIE), the Applied Computational Electromagnetics Society (ACES), the Progress In Electromagnetics Research (PIER) Electromagnetics Academy, and the National Academy of Inventors (NAI).

List of Contributors

Sensong An

Department of Electrical & Computer Engineering

University of Massachusetts Lowell

Lowell, MA

USA

and

Department of Materials Science & Engineering

Massachusetts Institute of Technology

Cambridge, MA

USA

N. Anselmi

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

Wenshan Cai

School of Electrical and Computer Engineering

Georgia Institute of Technology

Atlanta, GA

USA

Sawyer D. Campbell

The Pennsylvania State University

University Park, PA

USA

Yu Cao

School of Electrical, Computer and Energy Engineering

Arizona State University

Tempe, AZ

USA

Nagadastagiri Reddy Challapalle

School of Electrical Engineering and Computer Science

The Pennsylvania State University

University Park, PA

USA

Christos Christodoulou

Department of Electrical and Computer Engineering

The University of New Mexico

Albuquerque, NM

USA

Xiaocong Du

School of Electrical, Computer and Energy Engineering

Arizona State University

Tempe, AZ

USA

Ahmet M. Elbir

Interdisciplinary Centre for Security

Reliability and Trust (SnT)

University of Luxembourg

Luxembourg

Jonathan A. Fan

Department of Electrical Engineering

Stanford University

Stanford, CA

USA

Feng Feng

School of Microelectronics

Tianjin University

Tianjin

China

Clayton Fowler

Department of Electrical & Computer Engineering

University of Massachusetts Lowell

Lowell, MA

USA

Isha Garg

Elmore School of Electrical and Computer Engineering

Purdue University

West Lafayette, IN

USA

Arjun Gupta

Facebook

Menlo Park, CA

USA

Rong-Han Hong

Institute of Electromagnetics and Acoustics

Xiamen University

Xiamen

China

Wei Hong

State Key Laboratory of Millimeter Waves

School of Information Science and Engineering

Southeast University

Nanjing, Jiangsu Province

China

and

Department of New Communications

Purple Mountain Laboratories

Nanjing, Jiangsu Province

China

Hao-Jie Hu

Institute of Electromagnetics and Acoustics

Xiamen University

Xiamen

China

Ronald P. Jenkins

The Pennsylvania State University

University Park, PA

USA

Jing Jin

College of Physical Science and Technology

Central China Normal University

Wuhan

China

Youngeun Kim

School of Engineering & Applied Science

Yale University

New Haven, CT

USA

Youngwook Kim

Electronic Engineering

Sogang University

Seoul

South Korea

Slawomir Koziel

Faculty of Electronics, Telecommunications and Informatics

Gdansk University of Technology

Gdansk

Poland

and

Engineering Optimization & Modeling Center

Reykjavik University

Reykjavik

Iceland

Gokul Krishnan

School of Electrical, Computer and Energy Engineering

Arizona State University

Tempe, AZ

USA

Mrinal Kumar

Department of Mechanical and Aerospace Engineering

The Ohio State University

Columbus, OH

USA

Chonghan Lee

School of Electrical Engineering and Computer Science

The Pennsylvania State University

University Park, PA

USA

Yuhang Li

School of Engineering & Applied Science

Yale University

New Haven, CT

USA

Qing Liu

Institute of Electromagnetics and Acoustics

Xiamen University

Xiamen

China

Zhaocheng Liu

School of Electrical and Computer Engineering

Georgia Institute of Technology

Atlanta, GA

USA

Robert Lupoiu

Department of Electrical Engineering

Stanford University

Stanford, CA

USA

Jordan M. Malof

Department of Electrical and Computer Engineering

Duke University

Durham, NC

USA

Manel Martínez-Ramón

Department of Electrical and Computer Engineering

The University of New Mexico

Albuquerque, NM

USA

A. Massa

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

and

ELEDIA Research Center (ELEDIA@TSINGHUA – Tsinghua University)

Haidian, Beijing

China

and

ELEDIA Research Center (ELEDIA@UESTC – UESTC)

School of Electronic Science and Engineering University of Electronic Science and Technology of China

Chengdu

China

and

School of Electrical Engineering

Tel Aviv University

Tel Aviv

Israel

and

ELEDIA Research Center (ELEDIA@UIC – University of Illinois Chicago)

Chicago, IL

USA

Kumar Vijay Mishra

Computational and Information Sciences Directorate (CISD)

United States DEVCOM Army Research Laboratory

Adelphi, MD

USA

Weicong Na

Faculty of Information Technology

Beijing University of Technology

Beijing

China

Vijaykrishnan Narayanan

School of Electrical Engineering and Computer Science

The Pennsylvania State University

University Park, PA

USA

Indranil Nayak

ElectroScience Laboratory and Department of Electrical and Computer Engineering

The Ohio State University

Columbus, OH

USA

G. Oliveri

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

Willie J. Padilla

Department of Electrical and Computer Engineering

Duke University

Durham, NC

USA

Priyadarshini Panda

School of Engineering & Applied Science

Yale University

New Haven, CT

USA

Anna Pietrenko-Dabrowska

Faculty of Electronics, Telecommunications and Informatics

Gdansk University of Technology

Gdansk

Poland

L. Poli

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

A. Polo

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

Simiao Ren

Department of Electrical and Computer Engineering

Duke University

Durham, NC

USA

P. Rocca

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

and

ELEDIA Research Center (ELEDIA@XIDIAN – Xidian University)

Xi'an, Shaanxi Province

China

José Luis Rojo Álvarez

Departamento de Teoría de la señal y Comunicaciones y Sistemas Telemáticos y Computación

Universidad rey Juan Carlos

Fuenlabrada, Madrid

Spain

Kaushik Roy

Elmore School of Electrical and Computer Engineering

Purdue University

West Lafayette, IN

USA

M. Salucci

ELEDIA Research Center (ELEDIA@UniTN – University of Trento)

DICAM – Department of Civil, Environmental, and Mechanical Engineering

Trento

Italy

Wei Shao

School of Physics, University of Electronic Science and Technology of China

Institute of Applied Physics

Chengdu

China

Jingbo Sun

School of Electrical, Computer and Energy Engineering

Arizona State University

Tempe, AZ

USA

Fernando L. Teixeira

ElectroScience Laboratory and Department of Electrical and Computer Engineering

The Ohio State University

Columbus, OH

USA

Yeshwanth Venkatesha

School of Engineering & Applied Science

Yale University

New Haven, CT

USA

Bing-Zhong Wang

School of Physics, University of Electronic Science and Technology of China

Institute of Applied Physics

Chengdu

China

Haiming Wang

State Key Laboratory of Millimeter Waves

School of Information Science and Engineering

Southeast University

Nanjing, Jiangsu Province

China

and

Department of New Communications

Purple Mountain Laboratories

Nanjing, Jiangsu Province

China

Zhenyu Wang

School of Electrical, Computer and Energy Engineering

Arizona State University

Tempe, AZ

USA

Douglas H. Werner

The Pennsylvania State University

University Park, PA

USA

Qi Wu

State Key Laboratory of Millimeter Waves

School of Information Science and Engineering

Southeast University

Nanjing, Jiangsu Province

China

and

Department of New Communications

Purple Mountain Laboratories

Nanjing, Jiangsu Province

China

Li-Ye Xiao

Department of Electronic Science

Xiamen University, Institute of Electromagnetics and Acoustics

Xiamen

China

Amir I. Zaghloul

Bradley Department of Electrical and Computer Engineering

Virginia Tech

Blacksburg, VA

USA

Hualiang Zhang

Department of Electrical & Computer Engineering

University of Massachusetts Lowell

Lowell, MA

USA

Qi-Jun Zhang

Department of Electronics

Carleton University

Ottawa, ON

Canada

Bowen Zheng

Department of Electrical & Computer Engineering

University of Massachusetts Lowell

Lowell, MA

USA

Yi Zheng

School of Electrical Engineering and Computer Science

The Pennsylvania State University

University Park, PA

USA

Dayu Zhu

School of Electrical and Computer Engineering

Georgia Institute of Technology

Atlanta, GA

USA

Preface

The subject of this book is the application of the rapidly growing areas of artificial intelligence (AI) and deep learning (DL) in electromagnetics (EMs). AI and DL have the potential to disrupt the state-of-the-art in a number of research disciplines within the greater electromagnetics, optics, and photonics fields, particularly in the areas of inverse-modeling and inverse-design. While a number of high-profile papers have been published in these areas in the last few years, many researchers and engineers have yet to explore AI and DL solutions for their problems of interest. Nevertheless, the use of AI and DL within electromagnetics and other technical areas is only set to grow as more scientists and engineers learn about how to apply these techniques to their research. To this end, we organized this book to serve both as an introduction to the basics of AI and DL as well as to present cutting-edge research advances in applications of AI and DL in radio-frequency (RF) and optical modeling, simulation, and inverse-design. This book provides a comprehensive treatment of the field on subjects ranging from fundamental theoretical principles and new technological developments to state-of-the-art device design, as well as examples encompassing a wide range of related sub-areas. The content of the book covers all-dielectric and metallo-dielectric optical metasurface deep-learning-accelerated inverse-design, deep neural networks for inverse scattering and the inverse design of artificial electromagnetic materials, applications of deep learning for advanced antenna and array design, reduced-order model development, and other related topics.

This volume seeks to address questions such as “What is deep learning?,” “How does one train a deep neural network,?” “How does one apply AI/DL to electromagnetics, optics, scattering, and propagation problems?,” and “What is the current state-of-the-art in applied AI/DL in electromagnetics?” The first chapters of the book provide a comprehensive overview of the fundamental concepts and taxonomy of artificial intelligence, neural networks, and deep learning in order to provide the reader with a firm foundation on which to stand before exploring the more technical application areas presented in the remaining chapters. Throughout this volume, theoretical discussions are complemented by a broad range of design examples and numerical studies. We hope that this book will be an indispensable resource for graduate students, researchers, and professionals in the greater electromagnetics, antennas, photonics, and optical communities.

This book comprises a total of 17 invited chapters contributed from leading experts in the fields of AI, DL, computer science, optics, photonics, and electromagnetics. A brief summary of each chapter is provided as follows.

Chapter 1 introduces the fundamentals of neural networks and a taxonomy of terms, concepts, and language that is commonly used in AI and DL works. Moreover, the chapter contains a discussion of model development and how backpropagation is used to train complex network architectures. Chapter 2 provides a survey of recent advancements in AI and DL in the areas of supervised and unsupervised learning, physics-inspired machine learning models, among others as well as a discussion of the various types of hardware that is used to efficiently train neural networks. Chapter 3 focuses on the use of machine learning and surrogate models within the system-by-design paradigm for the efficient optimization-driven solution of complex electromagnetic design problems such as reflectarrays and metamaterial lenses. Chapter 4 introduces both the fundamentals and advanced formulations of artificial neural network (ANN) techniques for knowledge-based parametric electromagnetic (EM) modeling and optimization of microwave components. Chapter 5 presents two semi-supervised learning schemes to model microwave passive components for antenna and array modeling and optimization, and an autoencoder neural network used to reduce time-domain simulation data dimensionality. Chapter 6 introduces generative machine learning for photonic design which enables users to provide a desired transmittance profile to a trained deep neural network which then produces the structure which yields the desired spectra; a true inverse-design scheme. Chapter 7 discusses emergent concepts at the interface of the data sciences and conventional computational electromagnetics (CEM) algorithms (e.g. those based on finite differences, finite elements, and the method of moments). Chapter 8 combines DL with multiobjective optimization to examine the tradeoffs between performance and fabrication process uncertainties of nanofabricated optical metasurfaces with the goal of pushing optical metasurface fabrication toward wafer-scale. Chapter 9 explores machine learning (ML)/DL techniques to reduce the computational cost associated with the inverse-design of reconfigurable intelligent surfaces (RISs) which offer the potential for adaptable wireless channels and smart radio environments. Chapter 10 presents a selection of neural network architectures for Huygens' metasurface design (e.g. fully connected neural networks, convolutional neural networks, recurrent neural networks, and generative adversarial networks) while discussing neuromorphic photonics wherein meta-atoms can be used to physically construct neural networks for optical computing. Chapter 11 examines the use of deep neural networks in the design synthesis of artificial electromagnetic materials. For both forward and inverse design paradigms, the major fundamental challenges of design within that paradigm, and how deep neural networks have recently been used to overcome these challenges are presented. Chapter 12 introduces the framework of machine learning-assisted optimization (MLAO) and discusses its application to antenna and antenna array design as a way to overcome the limitations of traditional design methodologies. Chapter 13 summarizes the basics of uniform and non-uniform array processing using kernel learning methods which are naturally well adapted to the signal processing nature of antenna arrays. Chapter 14 describes a procedure for improved-efficacy electromagnetic-driven global optimization of high-frequency structures by exploiting response feature technology along with inverse surrogates to permit rapid determination of the parameter space components while rendering a high-quality starting point, which requires only further local refinement. Chapter 15 introduces four DL techniques to reduce the computational burden of high contrast inverse scattering of electrically large structures. These techniques can accelerate the process of reconstructing model parameters such as permittivity, conductivity, and permeability of unknown objects located inside an inaccessible region by analyzing the scattered fields from a domain of interest. Chapter 16 describes various applications of DL in the classification of radar images such as micro-Doppler spectrograms, range-Doppler diagrams, and synthetic aperture radar images for applications including human motion classification, hand gesture recognition, drone detection, vehicle detection, ship detection, and more. Finally, Chapter 17 explores the use of Koopman autoencoders for producing reduced-order models that mitigate the computational burden of traditional electromagnetic particle-in-cell algorithms, which are used to simulate kinetic plasmas due to their ability to accurately capture complicated transient nonlinear phenomena.

We owe a great debt to all of the authors of each of the 17 chapters for their wonderful contributions to this book, which we believe will provide readers with a timely and invaluable reference to the current state-of-the-art in applied AI and DL in electromagnetics. We would also like to express our gratitude to the Wiley/IEEE Press staff for their assistance and patience throughout the entire process of realizing this book – without their help, none of this would be possible.

June 2023

Sawyer D. Campbell and Douglas H. Werner 

Department of Electrical Engineering                

The Pennsylvania State University                      

University Park, Pennsylvania, USA                   

Section IIntroduction to AI-Based Regression and Classification

1Introduction to Neural Networks

Isha Garg and Kaushik Roy

Elmore School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA

The availability of compute power and abundance of data has resulted in the tremendous success of deep learning algorithms. Neural Networks often outperform their human counterparts in a variety of tasks, ranging from image classification to sentiment analysis of text. Neural networks can even play video games [1] and generate artwork [2]. In this chapter we introduce the basic concepts needed to understand neural networks. The simplest way to think of how these networks learn is to think of how a human learns to play a sport, let's say tennis. When someone who has never played tennis is put on a court, it takes them just a few volleys to figure out how to respond to an incoming shot. It might take a long time to get good at a sport, but it is quite magical that only through some trial and error, we can learn how to swing a racquet to a tennis ball heading our way. If we were to write a mathematical model to calculate the angle of swing, it would require many complicated variables such as the wind velocity, the incoming angle, the height from the ground, etc. Yet, we just learn from real-life examples that swinging a particular way has a particular impact, without knowing these variables. This implicit learning from examples is what sets machine learning models apart from their rule-based computational counterparts, such as a calculator. These models learn an implicit structure of the data they see without any explicit definitions on what to look for. The learning is guided by a lot of examples available with ground truth, and the models learn what is needed from these datapoints. Not only do they learn the datapoints they have seen, they are also able to generalize to unseen examples. For instance, we can train a model to differentiate between cats and dogs with, say, 100 examples. Now when we show them new examples of cats that were not present in the training set, they are still able to classify them correctly. There are enormous applications of the field, and a lot of ever-evolving subfields. In Section 1.1, we introduce some basic taxonomy of concepts that will help us understand the basics of neural networks.

1.1 Taxonomy

1.1.1 Supervised Versus Unsupervised Learning

In the unsupervised learning scenario, datapoints are present without labels. The aim is to learn an internal latent representation of data that catches repeated patterns and can make some decisions based on it. By latent representations, we mean an unexposed representation of data that is no longer in the original format, such as pixels of images. The hope is that repetition magnifies the significant aspects of images, or aids in learning compressed internal representations that can remove noisy artifacts or just learn lower dimensional representations for compressed storage, such as in AutoEncoders [3]. An advantage of having a meaningful latent space is that it can be used to generate new data. Most concepts covered in this chapter pertain to discriminatory models for regression or classification. However, in models such as Variational AutoEncoders [4], the latent space can be perturbed in order to create new data that is not present in the dataset. These models are called generative models. Unsupervised learning problems are harder and an active area of research, since it removes the need to label data. In this chapter, we will stick to supervised learning problems of the discriminatory kind.

1.1.2 Regression Versus Classification

The kind of output expected from a supervised learning discriminatory task dictates whether the problem is one of regression or classification. In regression, the output is a continuous value, such as predicting the price of a house. In classification, the task corresponds to figuring out which of the predefined classes an input belongs to. For example, determining whether an image is of a cat or a dog is a two-class classification problem. In this chapter, we will give examples of both regression and classification tasks.

1.1.3 Training, Validation, and Test Sets