B1674
Title: On using graph distances to estimate euclidean and related distances
Authors: Ery Arias-Castro - UC San Diego (United States) [presenting]
Abstract: Graph distances have proven quite useful in machine learning/statistics, particularly in the estimation of Euclidean or geodesic distances, and as such have been used to embed a graph (the multidimensional scaling problem). A partial review of the literature will be included and then more recent developments will be presented, including the minimax estimation of distances on a surface and consequences for manifold learning; the estimation of curvature-constrained distances on a surface; and the estimation of Euclidean distances based on an unweighted and noisy neighborhood graph.