Jian-Ming Jin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jian-Ming Jin has a total of 78 co-authors. Here are the top co-authors by number of shared publications:
- Su Yan
Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, 61801-2991, USA
12 shared publications - Hongliang Li
Resonant, Inc., San Mateo, CA, USA
7 shared publications - Kedi Zhang
Center for Computational Electromagnetics, Department of Electrical and Computer Engineering University of Illinois at Urbana‐Champaign Urbana IL 61801 USA
3 shared publications Zaiping Nie
Department of Microwave Engineering, University of Electronic Science and Technology of China, Chengdu, China
3 shared publications- Zhi Ning Chen
National University of Singapore
3 shared publications - Chao-Fu Wang
Nanjing University of Science and Technology
2 shared publications
Latest Publications
- Top Cited Papers
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (5) , 4051-4066
Deep Learning-Enabled Generative Antenna Design: Review and Prospects
This paper explores generative antenna design (GAD) and optimization using neural network (NN)-based deep learning (DL) methods. We first outline traditional antenna design and the challenges of applying NN-DL synthesis, particularly with generative adversarial networks (GANs). While conventional approaches reach performance limits, applying NN-DL generative methods to designs with increased degrees of freedom offers new possibilities for enhanced performance, additional functionality, and knowledge discovery. We then review the historical development and diverse applications of NN-based antenna optimization, covering individual elements, multi-antennas, and periodic structures. Key algorithmic concepts such as conditional GANs and variational autoencoders (VAEs) are discussed, emphasizing the role of prior knowledge in GAD. Recent advancements demonstrate how DL-enabled synthesis can surpass phase shift limits, control dispersion, and enhance metalens antenna gain bandwidth by 52%. The approach also enables full-range amplitude and phase control, effectively reducing sidelobe levels to below −30 dB. Finally, we highlight the significance of prior knowledge in GAD, distinguishing GAD from conventional parametric optimization. We discuss the challenges, future trends, and broader impacts of NN-driven generative design in next-generation antenna and electromagnetic device engineering.
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (5) , 4051-4066
Top Cited PapersDeep Learning-Enabled Generative Antenna Design: Review and Prospects
This paper explores generative antenna design (GAD) and optimization using neural network (NN)-based deep learning (DL) methods. We first outline traditional antenna design and the challenges of applying NN-DL synthesis, particularly with generative adversarial networks (GANs). While conventional approaches reach performance limits, applying NN-DL generative methods to designs with increased degrees of freedom offers new possibilities for enhanced performance, additional functionality, and knowledge discovery. We then review the historical development and diverse applications of NN-based antenna optimization, covering individual elements, multi-antennas, and periodic structures. Key algorithmic concepts such as conditional GANs and variational autoencoders (VAEs) are discussed, emphasizing the role of prior knowledge in GAD. Recent advancements demonstrate how DL-enabled synthesis can surpass phase shift limits, control dispersion, and enhance metalens antenna gain bandwidth by 52%. The approach also enables full-range amplitude and phase control, effectively reducing sidelobe levels to below −30 dB. Finally, we highlight the significance of prior knowledge in GAD, distinguishing GAD from conventional parametric optimization. We discuss the challenges, future trends, and broader impacts of NN-driven generative design in next-generation antenna and electromagnetic device engineering.
- 19 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 10, 28-37
Physics-Informed Machine Learning for the Efficient Modeling of High-Frequency Devices
In this paper, we present a machine learning technique based on analytic extension of eigenvalues and neural networks for the efficient modeling of high-frequency devices. In the proposed method, neural networks are used to learn the mapping between device's geometry and its modal equivalent circuit parameters. These circuit parameters are extracted from the eigen-decomposition of the deviceâs $Z$ -parameters at a few sample frequencies. The eigenvalues and eigenvectors of the $Z$ -matrix are analytically extended to other frequencies based on functional equations constructed from the lumped equivalent circuit model, from which the full electromagnetic response can be recovered. In addition to fully-connected neural network layers, our proposed model introduces an analytical projection branch based on AEE principles to maximize the information gain from samples in the training dataset. To improve the robustness and efficiency of the learning process, we introduce an adaptive gradient update algorithm. The overall model is end-to-end differentiable and can be integrated into gradient-based optimization methods. Numerical examples are provided to demonstrate the capability of the proposed method. Compared with traditional neural network-based models, the proposed approach achieves higher accuracy using fewer data samples and generalizes better to out-of-domain inputs.
- 19 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 10, 28-37
Physics-Informed Machine Learning for the Efficient Modeling of High-Frequency Devices
In this paper, we present a machine learning technique based on analytic extension of eigenvalues and neural networks for the efficient modeling of high-frequency devices. In the proposed method, neural networks are used to learn the mapping between device's geometry and its modal equivalent circuit parameters. These circuit parameters are extracted from the eigen-decomposition of the deviceâs $Z$ -parameters at a few sample frequencies. The eigenvalues and eigenvectors of the $Z$ -matrix are analytically extended to other frequencies based on functional equations constructed from the lumped equivalent circuit model, from which the full electromagnetic response can be recovered. In addition to fully-connected neural network layers, our proposed model introduces an analytical projection branch based on AEE principles to maximize the information gain from samples in the training dataset. To improve the robustness and efficiency of the learning process, we introduce an adaptive gradient update algorithm. The overall model is end-to-end differentiable and can be integrated into gradient-based optimization methods. Numerical examples are provided to demonstrate the capability of the proposed method. Compared with traditional neural network-based models, the proposed approach achieves higher accuracy using fewer data samples and generalizes better to out-of-domain inputs.
- Open Access
- 5 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 9, 157-165
A Hybrid Electromagnetic Optimization Method Based on Physics-Informed Machine Learning
In this paper, we present an optimization method based on the hybridization of the genetic algorithm (GA) and gradient optimization (grad-opt) and facilitated by a physicsinformed machine learning model. In the proposed method, the slow-but-global GA is used as a pre-screening tool to provide good initial values to the fast-but-local grad-opt. We introduce a robust metric to measure the goodness of the designs as starting points and use a set of control parameters to fine tune the optimization dynamics. We utilize the machine learning with analytic extension of eigenvalues (ML w/AEE) model to integrate the two pieces seamlessly and accelerate the optimization process by speeding up forward evaluation in GA and gradient calculation in grad-opt. We employ the divide-and-conquer strategy to further improve modeling efficiency and accelerate the design process and propose the use of a fusion module to allow for end-to-end gradient propagation. Two numerical examples are included to show the robustness and efficiency of the proposed method, compared with traditional approaches.
- 5 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 9, 157-165
Open AccessA Hybrid Electromagnetic Optimization Method Based on Physics-Informed Machine Learning
In this paper, we present an optimization method based on the hybridization of the genetic algorithm (GA) and gradient optimization (grad-opt) and facilitated by a physicsinformed machine learning model. In the proposed method, the slow-but-global GA is used as a pre-screening tool to provide good initial values to the fast-but-local grad-opt. We introduce a robust metric to measure the goodness of the designs as starting points and use a set of control parameters to fine tune the optimization dynamics. We utilize the machine learning with analytic extension of eigenvalues (ML w/AEE) model to integrate the two pieces seamlessly and accelerate the optimization process by speeding up forward evaluation in GA and gradient calculation in grad-opt. We employ the divide-and-conquer strategy to further improve modeling efficiency and accelerate the design process and propose the use of a fusion module to allow for end-to-end gradient propagation. Two numerical examples are included to show the robustness and efficiency of the proposed method, compared with traditional approaches.
- 17 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 70 (7) , 759-771
Efficient 3-D Finite Element Modeling of Periodic XBAR Resonators
An efficient technique is presented for 3D finite element modeling of large-scale periodic XBAR resonators in the time-harmonic domain. In this technique, a domain decomposition scheme is employed to decompose the computational domain into many small subdomains whose FE subsystems can be factorized with a direct sparse solver at a low cost. Transmission conditions are enforced to interconnect adjacent subdomains, and a global interface system is formulated and solved iteratively. To accelerate the convergence, a second-order transmission condition (SOTC) is designed to make the subdomain interfaces transparent for propagating and evanescent waves. An effective forward-backward preconditioner is constructed that, when combined with the SOTC, significantly reduce the number of iterations at no additional cost. Numerical results are given to demonstrate the accuracy, efficiency, and capability of the proposed algorithm.
- 17 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
- Vol. 70 (7) , 759-771
Efficient 3-D Finite Element Modeling of Periodic XBAR Resonators
An efficient technique is presented for 3D finite element modeling of large-scale periodic XBAR resonators in the time-harmonic domain. In this technique, a domain decomposition scheme is employed to decompose the computational domain into many small subdomains whose FE subsystems can be factorized with a direct sparse solver at a low cost. Transmission conditions are enforced to interconnect adjacent subdomains, and a global interface system is formulated and solved iteratively. To accelerate the convergence, a second-order transmission condition (SOTC) is designed to make the subdomain interfaces transparent for propagating and evanescent waves. An effective forward-backward preconditioner is constructed that, when combined with the SOTC, significantly reduce the number of iterations at no additional cost. Numerical results are given to demonstrate the accuracy, efficiency, and capability of the proposed algorithm.
- 16 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 8, 1-10
Electromagnetic-Thermal Modeling of Nonlinear Magnetic Materials
A nonlinear electromagnetic (EM)-thermal coupled solver is developed for modeling ferromagnetic materials widely used in electric motors. To accurately predict machine performance, the time-domain finite element method is employed to solve this multiphysics problem. By adopting the nonlinear B-H models to account for hysteresis effects, magnetic core losses are computed as the major sources of power dissipation for magnetic materials. The resulting temperature change is then obtained and its effect on the magnetic properties is subsequently evaluated. Due to different time scales of EM field variations and heat transfer processes, different time step sizes are adopted to enhance the simulation speed. During thermal time marching, the EM solver is invoked adaptively based on material property changes, and EM losses are calculated and updated through extrapolation, resulting in an efficient EM-thermal coupling scheme. Numerical examples are presented to validate the accuracy and capabilities of the proposed EM-thermal co-simulation framework.
- 16 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Multiscale and Multiphysics Computational Techniques
- Vol. 8, 1-10
Electromagnetic-Thermal Modeling of Nonlinear Magnetic Materials
A nonlinear electromagnetic (EM)-thermal coupled solver is developed for modeling ferromagnetic materials widely used in electric motors. To accurately predict machine performance, the time-domain finite element method is employed to solve this multiphysics problem. By adopting the nonlinear B-H models to account for hysteresis effects, magnetic core losses are computed as the major sources of power dissipation for magnetic materials. The resulting temperature change is then obtained and its effect on the magnetic properties is subsequently evaluated. Due to different time scales of EM field variations and heat transfer processes, different time step sizes are adopted to enhance the simulation speed. During thermal time marching, the EM solver is invoked adaptively based on material property changes, and EM losses are calculated and updated through extrapolation, resulting in an efficient EM-thermal coupling scheme. Numerical examples are presented to validate the accuracy and capabilities of the proposed EM-thermal co-simulation framework.
- 5 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 002
Efficient Broadband Modeling of Microwave Devices with Machine Learning and Analytical Extension of Eigenvalues
Accurate modeling of the broadband behavior of RF and microwave devices is of great interest as it can facilitate design optimization and high-level simulation. Machine learning (ML) techniques, in particular neural networks (NN), have recently gained recognition as a powerful tool for this purpose. Once trained, NNs can serve as an alternative to conventional methods such as full-wave simulations, which can be computationally expensive, or empirical models, which can have limited ranges of validity and accuracy. On the other hand, traditional NN-based modeling requires a large amount of training data and tends to have drastically worse performance on out-of-domain inputs. This coupled with the high computational cost for obtaining training data, severely hinders the application of NNs in device modeling.
- 5 September 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 002
Efficient Broadband Modeling of Microwave Devices with Machine Learning and Analytical Extension of Eigenvalues
Accurate modeling of the broadband behavior of RF and microwave devices is of great interest as it can facilitate design optimization and high-level simulation. Machine learning (ML) techniques, in particular neural networks (NN), have recently gained recognition as a powerful tool for this purpose. Once trained, NNs can serve as an alternative to conventional methods such as full-wave simulations, which can be computationally expensive, or empirical models, which can have limited ranges of validity and accuracy. On the other hand, traditional NN-based modeling requires a large amount of training data and tends to have drastically worse performance on out-of-domain inputs. This coupled with the high computational cost for obtaining training data, severely hinders the application of NNs in device modeling.
- 27 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 70 (7) , 4948-4952
Guest Editorial Machine Learning in Antenna Design, Modeling, and Measurements
Machine learning (ML) is the study of computational methods for improving performance by mechanizing the acquisition of knowledge from experience. As a modern data-driven optimization and applied regression methodology, ML aims to provide increasing levels of automation in the knowledge engineering process, replacing much time-consuming human activity with automatic techniques that improve accuracy and/or efficiency by discovering and exploiting regularities in training data. Indeed, many ML and data-driven methods, such as conventional artificial neural networks (ANNs), were introduced and studied within electromagnetics a few decades ago. However, these past studies did not benefit from the most recent advances in ML, which have been driven by the present confluence of improved hardware performance at lower cost, advanced network algorithms and architectures, data science, and considerable efforts dedicated to advancing the computational electromagnetics (CEM) benchmark. Today, a broader family of ML techniques based on ANNs has been developed. Examples include deep neural network (DNN), convolutional neural network (CNN), recurrent neural network, generative adversarial network, and deep reinforcement learning, which have been successfully applied to different engineering and science problems, ranging from image and video recognition, social media services, virtual personal assistant to autonomous vehicles, to name a few. This naturally suggests that applying ML to real-world electromagnetic problems could be one of the emerging trends in ML and artificial intelligence (AI) [1]–[4]. Indeed, ML has been becoming an important complement to existing experimental, computational, and theoretical aspects of electromagnetics.
- 27 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 70 (7) , 4948-4952
Guest Editorial Machine Learning in Antenna Design, Modeling, and Measurements
Machine learning (ML) is the study of computational methods for improving performance by mechanizing the acquisition of knowledge from experience. As a modern data-driven optimization and applied regression methodology, ML aims to provide increasing levels of automation in the knowledge engineering process, replacing much time-consuming human activity with automatic techniques that improve accuracy and/or efficiency by discovering and exploiting regularities in training data. Indeed, many ML and data-driven methods, such as conventional artificial neural networks (ANNs), were introduced and studied within electromagnetics a few decades ago. However, these past studies did not benefit from the most recent advances in ML, which have been driven by the present confluence of improved hardware performance at lower cost, advanced network algorithms and architectures, data science, and considerable efforts dedicated to advancing the computational electromagnetics (CEM) benchmark. Today, a broader family of ML techniques based on ANNs has been developed. Examples include deep neural network (DNN), convolutional neural network (CNN), recurrent neural network, generative adversarial network, and deep reinforcement learning, which have been successfully applied to different engineering and science problems, ranging from image and video recognition, social media services, virtual personal assistant to autonomous vehicles, to name a few. This naturally suggests that applying ML to real-world electromagnetic problems could be one of the emerging trends in ML and artificial intelligence (AI) [1]–[4]. Indeed, ML has been becoming an important complement to existing experimental, computational, and theoretical aspects of electromagnetics.
- Open Access
- 22 June 2022
- Published by Office of Scientific and Technical Information (OSTI)
Topology Optimization of Manufacturable Photonic Crystals With Complete Bandgaps [Dissertation]
The U.S. Department of Energy's Office of Scientific and Technical Information
- 22 June 2022
- Published by Office of Scientific and Technical Information (OSTI)
Open AccessTopology Optimization of Manufacturable Photonic Crystals With Complete Bandgaps [Dissertation]
The U.S. Department of Energy's Office of Scientific and Technical Information
- Open Access
- 15 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 69 (4) , 2078-2087
Second-Order Analytic Extension of Eigenvalues for Fast Frequency Sweep Analysis of RF Circuits
A second-order analytic extension of eigenvalue (AEE) method is presented and investigated for efficiently computing the Z-parameters of passive RF circuits over a wide frequency band with full-wave accuracy. The Z-parameters of an RF circuit are first extracted based on a full-wave simulation on sampling frequencies and then decomposed into eigenmodes, whose eigenvalues are analytically extended to all other frequencies within the frequency band of interest based on functional equations constructed from second-order series and parallel RLC circuits. An eigenvector-eigenvalue identity is adopted to compute the frequency-dependent eigenvectors from the eigenvalues of the submatrices, which are used in the expansion of the Z-parameters. A comparison with full-wave solutions is given for the second-order AEE where four frequencies are employed to accurately approximate the Z-parameters and predict the frequency response over the entire band of interest that goes to much higher frequencies than the first-order AEE. With this, the previously developed first-order AEE for a quasi-static analysis is successfully extended to much higher frequencies. Numerical examples are provided to validate the accuracy and demonstrate the capability of the second-order AEE. It is found that the second-order AEE is very accurate for modeling RF circuits with electrical sizes up to one wavelength that possibly contains resonances, as compared to the first-order AEE, which is applicable only to RF circuits with electrical sizes smaller than one-tenth to one-fifth of a wavelength that contains no resonance.
- 15 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 69 (4) , 2078-2087
Open AccessSecond-Order Analytic Extension of Eigenvalues for Fast Frequency Sweep Analysis of RF Circuits
A second-order analytic extension of eigenvalue (AEE) method is presented and investigated for efficiently computing the Z-parameters of passive RF circuits over a wide frequency band with full-wave accuracy. The Z-parameters of an RF circuit are first extracted based on a full-wave simulation on sampling frequencies and then decomposed into eigenmodes, whose eigenvalues are analytically extended to all other frequencies within the frequency band of interest based on functional equations constructed from second-order series and parallel RLC circuits. An eigenvector-eigenvalue identity is adopted to compute the frequency-dependent eigenvectors from the eigenvalues of the submatrices, which are used in the expansion of the Z-parameters. A comparison with full-wave solutions is given for the second-order AEE where four frequencies are employed to accurately approximate the Z-parameters and predict the frequency response over the entire band of interest that goes to much higher frequencies than the first-order AEE. With this, the previously developed first-order AEE for a quasi-static analysis is successfully extended to much higher frequencies. Numerical examples are provided to validate the accuracy and demonstrate the capability of the second-order AEE. It is found that the second-order AEE is very accurate for modeling RF circuits with electrical sizes up to one wavelength that possibly contains resonances, as compared to the first-order AEE, which is applicable only to RF circuits with electrical sizes smaller than one-tenth to one-fifth of a wavelength that contains no resonance.
- Open Access
- 28 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 69 (1) , 4-14
Fast Frequency Sweep Analysis of Passive Miniature RF Circuits Based on Analytic Extension of Eigenvalues
A fast frequency sweep approach based on analytic extension of eigenvalues (AEE) is presented and investigated for an efficient analysis of miniature passive RF circuits. In this approach, an eigenvalue decomposition is performed to the Z-parameters of the circuit components at one or a few frequencies, and analytic extension is applied to obtain the eigenvalues at all other frequencies within the band of interest. For electrically small circuit components, the frequency-independent characteristic inductances and capacitances, as well as the frequency-dependent resistances, are extracted from the eigenvalues at the sampling frequencies. The extracted characteristic parameters are then used to approximate the Z- and Y-parameters and predict the responses of the circuit over the entire frequency band. The accuracy of this approach is evaluated by comparing it with the results from a full-wave analysis. It is found that the proposed AEE is very accurate with a relative error of less than 1% for miniature RF circuits whose electrical sizes are smaller than one tenth of the wavelength and is more general and powerful than the one based on lumped equivalent circuits.
- 28 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Microwave Theory and Techniques
- Vol. 69 (1) , 4-14
Open AccessFast Frequency Sweep Analysis of Passive Miniature RF Circuits Based on Analytic Extension of Eigenvalues
A fast frequency sweep approach based on analytic extension of eigenvalues (AEE) is presented and investigated for an efficient analysis of miniature passive RF circuits. In this approach, an eigenvalue decomposition is performed to the Z-parameters of the circuit components at one or a few frequencies, and analytic extension is applied to obtain the eigenvalues at all other frequencies within the band of interest. For electrically small circuit components, the frequency-independent characteristic inductances and capacitances, as well as the frequency-dependent resistances, are extracted from the eigenvalues at the sampling frequencies. The extracted characteristic parameters are then used to approximate the Z- and Y-parameters and predict the responses of the circuit over the entire frequency band. The accuracy of this approach is evaluated by comparing it with the results from a full-wave analysis. It is found that the proposed AEE is very accurate with a relative error of less than 1% for miniature RF circuits whose electrical sizes are smaller than one tenth of the wavelength and is more general and powerful than the one based on lumped equivalent circuits.