This paper introduces a new latent variable probabilistic framework for representing spectral data of high spatial and spectral dimensionality, such as hyperspectral images. We use a generative ...