Convergence Rate of Frank-Wolfe for Non-Convex Objectives

07/01/2016

∙

We give a simple proof that the Frank-Wolfe algorithm obtains a stationary point at a rate of O(1/√(t)) on non-convex objectives with a Lipschitz continuous gradient. Our analysis is affine invariant and is the first, to the best of our knowledge, giving a similar rate to what was already proven for projected gradient methods (though on slightly different measures of stationarity).

READ FULL TEXT

Convergence Rate of Frank-Wolfe for Non-Convex Objectives

Convergence Error Analysis of Reflected Gradient Langevin Dynamics for Globally Optimizing Non-Convex Constrained Problems

Distributed Non-Convex First-Order Optimization and Information Processing: Lower Complexity Bounds and Rate Optimal Algorithms

A Gradient Smoothed Functional Algorithm with Truncated Cauchy Random Perturbations for Stochastic Optimization

Adaptive Gradient Methods Can Be Provably Faster than SGD after Finite Epochs

Improved SVRG for Non-Strongly-Convex or Sum-of-Non-Convex Objectives

Revisiting First-Order Convex Optimization Over Linear Spaces

AGGLIO: Global Optimization for Locally Convex Functions

Convergence Rate of Frank-Wolfe for Non-Convex Objectives

Related Research

Convergence Error Analysis of Reflected Gradient Langevin Dynamics for Globally Optimizing Non-Convex Constrained Problems

Distributed Non-Convex First-Order Optimization and Information Processing: Lower Complexity Bounds and Rate Optimal Algorithms

A Gradient Smoothed Functional Algorithm with Truncated Cauchy Random Perturbations for Stochastic Optimization

Adaptive Gradient Methods Can Be Provably Faster than SGD after Finite Epochs

Improved SVRG for Non-Strongly-Convex or Sum-of-Non-Convex Objectives

Revisiting First-Order Convex Optimization Over Linear Spaces

AGGLIO: Global Optimization for Locally Convex Functions