Low-Cost Multi-Objective Optimization of Multiparameter Antenna Structures Based on the l1 Optimization BPNN Surrogate Model

Abstract

The development of modern wireless communication systems not only requires the antenna to be lightweight, low cost, easy to manufacture and easy to integrate but also imposes requirements on the miniaturization, wideband, and multiband design of the antenna. Therefore, designing an antenna that quickly and effectively meets multiple performance requirements is of great significance. To solve the problem of the large computational cost of traditional multi-objective antenna design methods, this paper proposes a backpropagation neural network surrogate model based on l1 optimization (l1-BPNN). The l1 optimization method tends to punish larger weight values and select smaller weight values so as to preserve a small amount of important weights and reset relatively unimportant weights to zero. By using l1 optimization method, the network mapping structure can be automatically adjusted to achieve the most suitable and compact structure of the surrogate model. Furthermore, for multi-parameter antenna design problems, a fast multi-objective optimization framework is constructed using the proposed l1-BPNN as a surrogate model. The framework is illustrated using a miniaturized multiband antenna design case, and a comparison with previously published methods, as well as numerical validation, is also provided.

Cunninghamia lanceolata Canopy Relative Chlorophyll Content Estimation Based on Unmanned Aerial Vehicle Multispectral Imagery and Terrain Suitability Analysis

“…BPNN, as one of the most widely used neural networks, has a powerful nonlinear mapping ability, which allows it to approximate any complex nonlinear function through a multilayer network structure. This makes BP neural networks have higher accuracy and stronger generalization ability when dealing with data with complex nonlinear relationships [58]. However, in the construction of the BPNN model, too many input nodes not only increase the training time cost of the model, but also may cause the neural network to be too sensitive to the noise and redundant information in the input vectors, which reduces the generalization ability and prediction accuracy of the model.…”

Length-Dependent Deep Neural Network Based Modeling for High-Speed Channels

“…In addition, optimization procedures widely employ active-learning algorithms such as Bayesian optimization [9], [10], [11]. Alternatively, supervised and unsupervised learning are extensively exploited to enhance design efficiency using machine learning-based surrogate models [19], [20], [21], [22], [26], [28], [27], [16], [17], [18], [23], [24], [25], [12], [13], [14], [15], [29], [30], [31]. To this end, contemporary deep learning techniques with breakthroughs in various applications, such as deep neural networks (DNNs), have emerged as candidates for this surrogate model.…”

An efficient application of machine learning for designing user specific transmissive radome

“…Simulators using voltage-current distribution along the nodes of mesh of the structure(s) are highly accurate owing to full wave rigorous analysis (MoM, FEM, FDTD) but costly in computational efficiency. To minimize this cost, surrogate modeling (an engineering method opted when direct outcome is difficult to measure) has been explored in the last decade to provide solutions for scattering parameters, [12][13][14][15] reflect-array antennas, [16][17][18] and can be trained with forward modeling, or inverse modeling. 19,20 Considering the nonlinear relationship of the EM problem(s), machine learning (ML) stands out as an optimal solution for surrogate modeling.…”