Speaker: Martin Ehler (NIH/UMD)

Title: Arbitrary Shrinkage Rules for Approximation Schemes with Sparsity Constraints


Finding a sparse representation of a possibly noisy signal is a common problem in signal representation and processing. It can be modeled as a variational minimization with ℓτ-sparsity constraints for τ < 1. Applications whose computation time is crucial require fast algorithms for this minimization. However, there are no fast methods for finding the exact minimizer, and to circumvent this limitation, we consider minimization up to a constant factor. We verify that arbitrary shrinkage rules provide closed formulas for such minimizers, and we introduce a new shrinkage strategy, which is adapted to τ < 1.