Charles Explorer logo
🇬🇧

Hybrid LSQR regularization for inverse problems in Single Particle Analysis

Publication at Faculty of Mathematics and Physics |
2020

Abstract

In this contribution we concentrate on discrete inverse problems Ax b arising incryo-electron microscopy single particle analysis. Since these problems have spe-cific properties (such as high sensitivity of the solution on noise present in theobservation b, large noise level, etc.), their numerical solution is highly challen-ging.

We describe a variant of iterative hybrid LSQR method with inner Tikhonovregularization and show its effectivity on these problems. Special attention is gi-ven to parameter-choice method for the inner Tikhonov regularization as well assuitable stopping criterion for the outer LSQR iterations.

Since the underlying ap-plication belongs to the category of large scale problems, issues related to effectiveimplementation are also studied. Based on numerical experiments, we examine thebehavior of the proposed method on synthetic and real datasets.

Finaly, severalopen questions are formulated.