Ninja optimization algorithm based ultra wideband antenna electromagnetic band gap modeling via a generative adversarial network

Abstract

Ultra-wideband antennas with electromagnetic band-gap (EBG) structures play a crucial role in next-generation wireless and energy-efficient communication systems due to their ability to provide broad spectral coverage, high gain, and reduced interference. This paper presents an intelligent prediction framework that integrates a Generative Adversarial Network (GAN) with the Ninja Optimization Algorithm (NOA) to accurately model and predict the electromagnetic performance of ultra-wideband antenna-EBG configurations. The core innovation of the proposed framework lies in coupling adversarial learning with NOA-based optimization to enhance surrogate modeling accuracy and robustness for antenna-EBG systems. The proposed method is compared with multiple deep learning architectures, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN), and Artificial Neural Network (ANN) models. Experimental results demonstrate that the GAN tuned with the NOA achieves superior predictive accuracy, yielding a mean squared error of [Formula: see text], a root mean squared error of [Formula: see text], and a coefficient of determination of [Formula: see text]. The integration of the Ninja Optimization Algorithm significantly enhances learning stability, convergence rate, and generalization performance. The framework also demonstrates competitive robustness when benchmarked against hybrid optimization strategies such as PSO-GAN, BA-GAN, and DE-GAN. Overall, the proposed NOA-enhanced GAN establishes an efficient, scalable, and high-precision modeling pathway for the design and optimization of ultra-wideband antenna EBG structures, contributing to the advancement of intelligent communication and renewable energy systems.