A Disentangling Invertible Interpretation Network for Explaining Latent Representations | IEEE Conference Publication | IEEE Xplore