Self-supervised Hypergraphs for Learning Multiple World Interpretations | IEEE Conference Publication | IEEE Xplore