Bayesian Reimaging of Sparsity in Inverse Problems - IMAGINE Seminar Series

Attend a virtual seminar on Bayesian approaches to sparse inverse problems in imaging and dictionary learning. Explore how sparsity can be interpreted as an a priori belief about solutions in the Bayesian framework, and learn about efficient computational methods for Bayesian sparse inverse solvers. Discover how hierarchical prior models can be designed to promote sparsity in computed solutions, with parameters set according to expected sparsity levels and data sensitivity. Examine an inner-outer iteration scheme for solving weighted linear least squares problems and dynamically updating scaling weights. Understand how the Conjugate Gradient method for least squares (CGLS) can be applied to achieve automatic model reduction. View computed examples demonstrating the performance of this Bayesian approach in imaging and dictionary learning applications. Join the global scientific community in this OneWorld SIAM-IS virtual seminar series, providing a forum for idea exchange and networking in mathematical imaging and applied inverse problems during the COVID-19 pandemic.