Stability and minimax optimality of tangential Delaunay complexes for manifold reconstruction (Q1650797)

The authors look at reconstructed manifolds from points embedded in Euclidean space. This method requires a dense sample of the points from the manifold embedding and creates a \(k\)-dimensional approximation to the manifold. Extending this method, the authors create an estimator for the manifold that they show possesses ambient isotopy and Hausdorff close to the original manifold. This is based on a combined approach of tangential Delaunay complex and local principal component analysis. In addition, they show that the process still works with outliers provided a decluttering process is undertaken.