Through this work we explore the unsupervised topological learning of multimodal data presenting a complex structure allowing to learn their representations. We are particularly interested in heterogeneous data whose ...