Speaker: Hiroko Yamano

Abstract:
Real-world systems often include higher-order structures involving three or more units, which traditional pairwise network models fail to capture. For example, in human musculoskeletal networks, the connections between muscles and bones typically represent a many-to-many relationship, enabling complex organizational control of motor co-ordination with multiple degrees of freedom. However, despite extensive anatomical knowledge, this complex relationships among body components increase indeterminacy in coordination, leaving a systematic understanding of organizational control largely unknown. Here, we tackled this problem with hypergraph representation to model musculoskeletal system. We used both pairwise and hypergraph-based embedding methods to learn the connectivity of muscles. Experiments demonstrated the superiority of the proposed hypergraph-based method over pairwise methods in distinguishing the specific roles of the muscles connecting different body parts.