A PROXIMAL METHOD FOR COMPOSITE MINIMIZATION
A.S. LEWIS (ORIE Cornell)
Many important optimization problems can be formulated as minimizing a composite function h(c(.)), where the outer function h is convex but nonsmooth, and the inner function c is smooth. For example, a variety of statistical procedures, such as LASSO, LARS, and basis pursuit for sparse signal recovery, involve minimizing the sum of a smooth convex function and the L1 norm. In large-scale examples, subproblems formed by linearizing the inner function c and adding a quadratic regularizing term are often easy to solve. I describe an algorithm based on this idea.
Joint work with S.J. Wright (Wisconsin)