Georgia Institute of Technology College of Sciences

 There has been much recent work applying splitting algorithms to  optimization problems designed to produce sparse solutions. In this talk,  we'll look at extensions of these methods to the nonconvex case, motivated  by results in compressive sensing showing that nonconvex optimization can recover signals from many fewer measurements than convex optimization. Our examples of the application of these methods will include image reconstruction from few measurements, and the decomposition of high-dimensional datasets, most notably video, into low-dimensional and sparse components.