Using the Rips complex for topologically-certified manifold learning

by David Salinas (Gipsa-Lab Grenoble)
May 2nd at 2pm in Byron blanc

Given a set of points that sample a manifold, it has been  proved recently that, under some density requirements, the Rips complex  of the points provides an approximation of the manifold with the correct  topology.  One advantage of the Rips complex is that it is determined entirely by  its graph which provides a compact form of storage. Unfortunately, the  dimension of the Rips complex can be huge, compared to the intrinsic  dimension of the shape.  We will describe different strategies for simplifying the Rips complex  without changing its topology and obtain a simplicial complex whose  dimension is that of the sampled manifold.

 

Comments are closed.