Topological Deep Learning

Topological deep learning (TDL) aims to improve deep learning models by incorporating topological structures—like simplicial complexes and cell complexes—which represent higher-order relationships in data beyond simple pairwise connections found in graphs. Current research focuses on developing novel neural network architectures, such as simplicial neural networks and message-passing networks operating on these higher-order structures, and on creating benchmark datasets to evaluate their performance across various tasks. TDL's ability to capture complex interactions holds significant promise for applications in diverse fields, including material science, molecular property prediction, and analysis of complex networks.