Abstract: Benefitting from superior performance of learning a low-dimensional sparse representation, autoencoders (AEs) are widely applied to hyperspectral unmixing (HU) task, which aims at ...