Groups and interactions in data, networks and biology

Beyond Scalar Affinities for Network Analysis

Amit Singer

Princeton University
[SLIDES]

The popular representation of datasets and networks as undirected weighted graphs is instrumental for many purposes such as clustering, dimensionality reduction, semi-supervised learning, and more. In this talk we will explore several instances in which endowing the edges with vectors or transformations is beneficial for certain applications in computer vision, structural biology, and manifold learning.