Bidirectional Long Short-Term Memory-Based Amplitude and Phase Prediction of Antenna Array

This letter proposes a bidirectional long short-term memory (BiLSTM) network to predict the excitations of antenna arrays for arbitrary radiation patterns. Compared to artificial neural networks (ANN) and unidirectional long short-term memory (LSTM) networks, BiLSTM effectively captures phase, amplitude, and inter-element dependencies by processing sequences in forward and backward directions. Extensive evaluations demonstrate that BiLSTM surpasses ANN and LSTM in accuracy. These findings highlight BiLSTM's potential to significantly improve electromagnetic field analysis and advance practical applications in antenna design and measurement.