"Deep Learning-Based Phaseless Near-Field to Near-Field Transformation via Joint Height-Frequency Embedding"

"This dataset accompanies the paper entitled \u201cDeep Learning-Based Phaseless Near-Field to Near-Field Transformation via Joint Height-Frequency Embedding.\u201d It is designed for the training and evaluation of the proposed Joint Height-Frequency Embedding Network (JHFE-Net). Each sample contains an input magnetic near-field magnitude pattern at a height of 10 mm and a frequency of 3 GHz, together with a target near-field pattern corresponding to one of six heights ranging from 5 mm to 15 mm and one of three frequencies: 2 GHz, 3 GHz, and 4 GHz. The dataset was analytically generated using randomly distributed tangential magnetic dipoles and vertical electric dipoles with randomized positions, amplitudes, and phases. The scanning area is 200 mm \u00d7 200 mm with a sampling interval of 5 mm, resulting in 41 \u00d7 41 scanning points. The dataset can be used for phaseless near-field transformation, electromagnetic-field reconstruction, and deep-learning-based near-field analysis."