Poster
CCCP is Frank-Wolfe in disguise
Alp Yurtsever · Suvrit Sra
Hall J (level 1) #935
Keywords: [ convex-concave procedure ] [ frank-wolfe ] [ cccp ] [ expectation maximization ] [ difference of convex programming ] [ conditional gradient method ] [ sinkhorn ]
This paper uncovers a simple but rather surprising connection: it shows that the well-known convex-concave procedure (CCCP) and its generalization to constrained problems are both special cases of the Frank-Wolfe (FW) method. This connection not only provides insight of deep (in our opinion) pedagogical value, but also transfers the recently discovered convergence theory of nonconvex Frank-Wolfe methods immediately to CCCP, closing a long-standing gap in its non-asymptotic convergence theory. We hope the viewpoint uncovered by this paper spurs the transfer of other advances made for FW to both CCCP and its generalizations.