Convex Optimization without Projection Steps

TL;DR

This paper introduces a projection-free iterative algorithm for convex optimization over compact domains, extending Frank & Wolfe's method to arbitrary convex sets, with guarantees on convergence and solution sparsity.

Contribution

It generalizes the Frank & Wolfe algorithm to arbitrary convex domains, providing convergence guarantees and analyzing sparsity bounds for solutions in various convex optimization problems.

Findings

Guarantees {psilon}-small duality gap after O(1/{psilon}) iterations.

Establishes matching upper and lower bounds of psilon for solution sparsity.

Demonstrates practical efficiency for large matrix problems like matrix completion.

Abstract

For the general problem of minimizing a convex function over a compact convex domain, we will investigate a simple iterative approximation algorithm based on the method by Frank & Wolfe 1956, that does not need projection steps in order to stay inside the optimization domain. Instead of a projection step, the linearized problem defined by a current subgradient is solved, which gives a step direction that will naturally stay in the domain. Our framework generalizes the sparse greedy algorithm of Frank & Wolfe and its primal-dual analysis by Clarkson 2010 (and the low-rank SDP approach by Hazan 2008) to arbitrary convex domains. We give a convergence proof guaranteeing {\epsilon}-small duality gap after O(1/{\epsilon}) iterations. The method allows us to understand the sparsity of approximate solutions for any l1-regularized convex optimization problem (and for optimization over the simplex), expressed as a function of the approximation quality. We obtain matching upper and lower bounds of {\Theta}(1/{\epsilon}) for the sparsity for l1-problems. The same bounds apply to low-rank semidefinite optimization with bounded trace, showing that rank O(1/{\epsilon}) is best possible here as well. As another application, we obtain sparse matrices of O(1/{\epsilon}) non-zero entries as {\epsilon}-approximate solutions when optimizing any convex function over a class of diagonally dominant symmetric matrices. We show that our proposed first-order method also applies to nuclear norm and max-norm matrix optimization problems. For nuclear norm regularized optimization, such as matrix completion and low-rank recovery, we demonstrate the practical efficiency and scalability of our algorithm for large matrix problems, as e.g. the Netflix dataset. For general convex optimization over bounded matrix max-norm, our algorithm is the first with a convergence guarantee, to the best of our knowledge.