Eigenmatrix for Unstructured Sparse Recovery

Explore a colloquium talk on "Eigenmatrix for Unstructured Sparse Recovery" presented by Lexing Ying from Stanford University. Delve into the challenges of recovering spike locations and weights of unknown sparse signals from unstructured observations. Examine various applications, including rational approximation, spectral function estimation, Fourier inversion, Laplace inversion, and sparse deconvolution. Learn about the innovative eigenmatrix construction, a data-driven approach that transforms non-linear inverse problems into eigen-decomposition problems. Discover how this method extends the classical Prony's method and offers a novel solution for sparse, unstructured recovery problems. Gain insights into handling noise in sample values and the unstructured nature of sample locations throughout this 48-minute presentation from the University of Chicago Department of Mathematics colloquium series.