Active Subsampling Using Deep Generative Models by Maximizing Expected Information Gain | IEEE Conference Publication | IEEE Xplore