Feature‐Based Rips Complex (Graph to Simplicial)

This is a generalization of the usual Rips-Complex, where the pairwise distances can include a factor based on the Euclidean distance of the feature vectors. This may be useful in cases where the features give additional information about the underlying topology (e.g. two points appear close when embedded in Euclidean space but are far apart on the manifold). It can also be used to create models that have more expressive message-passing dynamics, with message passing being done over both the position-based topology and the feature-based topology.

From https://github.com/pyt-team/challenge-icml-2024/pull/35