An Away-Step Frank-Wolfe Method for Minimizing Logarithmically-Homogeneous Barriers

MaRDI QIDQ6438277zbMATH OpenFDO

Authors Renbo Zhao

Publication date 28 May 2023

Abstract: We present and analyze a new away-step Frank-Wolfe method for the convex optimization problem $mi{n}_{xinmathcalX};f(mathsfAx)+langlec,xangle$, where $f$ is a $heta$-logarithmically-homogeneous self-concordant barrier, $mathsfA$ is a linear operator, $langlec,cdotangle$ is a linear function and $mathcalX$ is a nonempty polytope. We establish affine-invariant global linear convergence rates for both the objective gaps and the Frank-Wolfe gaps generated by our method. When specialized to the D-optimal design problem, our results settle a question left open since Ahipasaoglu, Sun and Todd (2008). We also show that the iterates generated by our method will land on a face of $mathcalX$ in a finite number of iterations, and hence our method may have improved local linear convergence rates.

This page was built for publication: An Away-Step Frank-Wolfe Method for Minimizing Logarithmically-Homogeneous Barriers

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6438277)