Abstract: The problem in the classification of hyperspectral images has been tough since the spectral dimension is high, the number of labeled sample data is low, and there are computing limitations.