Abstract: Deep learning has shown great promise in hyperspectral unmixing (HU), especially the unmixing methods based on autoencoder (AE) networks, which are the most prevalent these days. Since most ...