Optimization of Antenna Design Using the Artificial Neural Network and the Simulated Annealing Algorithm

Jin-hua, Huang; Li, Wenting; He, Yejun; Zhang, Long; Wong, Sai‐Wai

doi:10.1109/comcomap53641.2021.9653055

Cited by 7 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Simulated annealing simulates the thermal equilibrium process in solid materials and leverages the similarity between the problem of random search optimization to find global optimal or approximately global optimal solutions. Some researchers have also proposed using genetic algorithms for optimizing neural networks [13], using simulated annealing to find optimal solutions [14] and even combining both approaches [15]. In practical applications, genetic algorithms are prone to premature convergence problems.…”

Section: Related Workmentioning

confidence: 99%

Two-stream neural network with different gradient update strategies

Zhang,

Luo,

Hao

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Deep neural networks will be affected by various noises in different scenes. Traditional deep neural networks often use gradient descent algorithms to update parameter weights. When the gradient falls to a certain range, it is easy to fall into the local optimal solution. Although the impulse method and other methods can escape from local optimization in some scenarios, they still have some limitations, which will greatly reduce the application effect of the actual scenes. To solve the above problems, a two-stream neural network with different gradient update strategies was proposed. Combined with the gradient ascent algorithm, this method alleviated the disadvantage of deep neural networks falling into local optimality and increased the robustness of neural networks to a certain extent. The experimental results on the CIFAR10 dataset verify that the proposed method can improve the accuracy of various gradient descent optimizers by about 1%, such as SGD, Adagrad, RM-Sprop and Adam. The experimental results on the COCO dataset show that the accuracy of the proposed method is also improved compared with the baseline models PAA and EfficientDet. The method proposed can be widely used in various neural network structures and has good practical significance and application prospects.

show abstract

Section: Related Workmentioning

confidence: 99%

Two-stream neural network with different gradient update strategies

Zhang,

Luo,

Hao

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…For instance, [23] replaced EM solvers with trained ML models such as Lasso, artificial neural network (ANN), and k-nearest neighbor (kNN) models, with a training data size of 450. In [24], an ANN was applied with a training data size of 110. These studies have demonstrated promising results in reduced time.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning-based Optimization of Hexagon-shaped Fractal Antenna for Ultra-wideband Communications

Indharapu,

Caruso,

Fields

et al. 2024

PIER C

View full text Add to dashboard Cite

In the wireless communication industry, achieving gigabit-per-second data rates with low-profile, ultra-wideband (UWB) microstrip patch antennae poses a significant challenge. Conventional optimization algorithms, though effective, are often computationally expensive, particularly for complex antenna geometries with high degrees of freedom. There is an imperative need for new methodologies to address this challenge and revolutionize the antenna optimization process. Successful and timely development of antennas relies on the efficiency and computational speed of optimization algorithms, full-wave electromagnetic solvers, and the intuition of radio frequency engineers. To mitigate the dependence on complex and time-consuming processes, we propose an efficient machine learning (ML)-based antenna optimization methodology that minimizes optimization time by more than 90%. This paper aims to apply and study the performance of two specific ML models, the radial basis function (RBF), and the least squared regression (LSR) models, in the bandwidth optimization without increasing the aperture area of a hexagon-shaped fractal antenna. The hexagon-shaped fractal antenna was chosen for its UWB characteristics, low profile, and high degrees of freedom (10 adjustable parameters). The reflection coefficient response of a hexagon-shaped fractal antenna is predicted by the trained RBF and LSR models and further optimized by the genetic algorithm (GA). The proposed approach stands out among other notable works in this research domain, especially for UWB applications, by prioritizing the optimization of the mean of the reflection coefficient across the entire frequency range instead of solely targeting individual frequency points. The GA-based optimization using trained ML models has increased the bandwidth by 21.3% and reduced the computational time by 90% compared to conventional optimization without increasing the physical or electrical size of the antenna. Simulation and measurement results concurred with a maximum difference of 5%, demonstrating the efficacy of the ML approach for antenna optimization.

show abstract

“…In recent years, artificial neural networks (ANNs) are rapidly emerging as a powerful tool in EM-based modeling and optimization [10], such as passive circuits and components [11][12][13], field-effect transistors [14], antennas [15][16][17][18], and fault detection and diagnosis [19], [20]. Deep learning (DL) enhanced the ability of ANNs by utilizing deeper networks and more neurons and has become a dominant paradigm in addressing more complex problems [21], [22].…”

mentioning

confidence: 99%

Fully Automated Design Method Based on Reinforcement Learning and Surrogate Modeling for Antenna Array Decoupling

Wei

Zhou

Wang

et al. 2023

IEEE Trans. Antennas Propagat.

View full text Add to dashboard Cite

Modern electromagnetic (EM) device design generally relies on extensive iterative optimizations by designers using simulation software (e.g. CST), which is a very timeconsuming and tedious process. To relieve human engineers and boost productivity, we proposed a machine learning framework to solve the problem of automated design for EM tasks. The proposed approach combines advanced reinforcement learning (RL) algorithms and deep neural networks (DNNs) in an attempt to simulate the decision-making process of human designers to realize automation learning. Specifically, the RL-based agent can interact with the EM design software without engaging human designers, allowing for automated design. Besides, the data accumulated during EM software simulation in the early design stage are reused as training data to build a DNN surrogate model to replace the time-consuming EM simulation and further accelerate the training of RL to achieve better optimization of EM design. Two types of antenna array decoupling including 1×2 and 1×4 arrays working at 3.5 GHz are used as test vehicles to validate the proposed method. The decoupling metasurfaces designed by the proposed fully automated method based on RL showed satisfactory results comparable to the results achievable by human designers. This indicates that the proposed method can be used to build powerful tools to boost the design efficiency of EM devices.

show abstract

Optimization of Antenna Design Using the Artificial Neural Network and the Simulated Annealing Algorithm

Cited by 7 publications

References 10 publications

Two-stream neural network with different gradient update strategies

Two-stream neural network with different gradient update strategies

Machine Learning-based Optimization of Hexagon-shaped Fractal Antenna for Ultra-wideband Communications

Fully Automated Design Method Based on Reinforcement Learning and Surrogate Modeling for Antenna Array Decoupling

Contact Info

Product

Resources

About