Graph reconstruction from random-walk co-visitation: Geometric, empirical, and controlled networks

概要

Reconstructing an unknown graph from the trajectory of a random walk arises both for spatial correlation networks in astrophysics and for connectivity inference in network science. We present a reconstruction pipeline whose observable is the random-walk co-visitation matrix, whose model is a pairwise edge-weight basis, and whose fitter is a frame-balanced Levenberg-Marquardt (fbLM) scheme with per-node group weights and a self-calibrated edge readout. Unlike the marginal occupation, the co-visit…