Scientists at HRL Laboratories have published their new framework for training computer deep neural networks to be able to classify synthetic aperture radar (SAR) images without a large labeled data set, solving the problem of SAR image identification when only a few labeled data were available.