Machine Learning and Inverse Problems
Many challenging image processing tasks can be described by an ill-posed linear inverse problem: deblurring, deconvolution, inpainting, compressed sensing, and superresolution all lie in this framework. Recent advances in machine learning and image processing have illustrated that it is often possible to learn a regularizer from training data that can outperform more traditional approaches by large margins. In this talk, I will describe the central prevailing themes and tradeoffs of this emerging area. The core ideas underlying these methods draw upon ideas from statistics, optimization, signal processing, and machine learning. We will explore a new class of approaches based on implicit “infinite-depth” networks that yield reconstruction frameworks with provable convergence guarantees and significant empirical performance gains. Finally, I will describe mechanisms for “model adaptation” — that is, given a model trained to solve one inverse problem with a known forward model, we propose novel procedures that adapt the network to a perturbed forward model. This is joint work with Davis Gilton and Greg Ongie.
Monday, September 20
1:25 pm MST/AZ
Zoom meeting room link: https://asu.zoom.us/j/6871076660
Note: The RTG Seminar will meet via Zoom, at least for the first few weeks of the semester. Depending on developments, we may hold some in-person meetings later in the term.