Traditional supervised deep learning (DL) methods for hyperspectral image (HSI) classification are severely limited by the quality and quantity of labels. Furthermore, existing feature extraction ...