Abstract: Spectrum prediction is challenging due to its multi-dimension, complex inherent dependency, and heterogeneity among the spectrum data. In this paper, we first propose a stacked autoencoder ...