Generative Adversarial Symmetry Discovery
Jianke Yang, Robin Walters, et al.
ICML 2023
The equivalence in exact arithmetic of the Lanczos tridiagonalization procedure and the conjugate gradient optimization procedure for solving Ax = b, where A is a real symmetric, positive definite matrix, is well known. We demonstrate that a relaxed equivalence is valid in the presence of errors. Specifically we demonstrate that local ε-orthonormality of the Lanczos vectors guarantees local ε-A-conjugacy of the direction vectors in the associated conjugate gradient procedure. Moreover we demonstrate that all the conjugate gradient relationships are satisfied approximately. Therefore, any statements valid for the conjugate gradient optimization procedure, which we show converges under very weak conditions, apply directly to the Lanczos procedure. We then use this equivalence to obtain an explanation of the Lanczos phenomenon: the empirically observed "convergence" of Lanczos eigenvalue procedures despite total loss of the global orthogonality of the Lanczos vectors. © 1980.
Jianke Yang, Robin Walters, et al.
ICML 2023
Kenneth L. Clarkson, K. Georg Hampel, et al.
VTC Spring 2007
Charles Micchelli
Journal of Approximation Theory
George Markowsky
J. Math. Anal. Appl.