Array Antenna Diagnosis Using Self-Paced Learning Convolutional Neural Networks

The array antenna diagnosis is known as a computationally-cost task that requires many iterations to get the solution. In this article, a self-paced learning convolutional neural network (CNN) is proposed for array diagnosis. The proposed method is able to efficiently detect the impaired elements from the measured far-field radiation patterns. However, due to the noise in the complex electromagnetic environment, the diagnostic accuracy of the method is significantly decreased. To address the problem, the self-paced learning strategy is utilized to enable CNN to gradually learn the samples from “easy” to “difficult”. As the training progresses, the CNN gradually improves its ability to learn difficult samples, becoming more adept at processing noisy samples. The proposed method is available for the planar array with mutual coupling effect. Numerical simulation are carried out to demonstrate the superiority of the proposed <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">me</sup>thod by comparing with the CNN method and the compressed sensing method.