1: \begin{abstract}

2: 		The Frank-Wolfe (FW) method, which implements efficient linear oracles that minimize linear approximations of the objective function over a {\it fixed} compact convex set, has recently received much attention in the optimization and machine learning literature. In this paper, we propose a new FW-type method for minimizing a smooth function over a compact set defined {\color{blue}as the level set of a single {\em difference-of-convex} function}, based on new {\em generalized} linear-optimization oracles (LO). We show that these LOs can be computed efficiently with {\it closed-form solutions} in some important optimization models that arise in compressed sensing and machine learning. In addition, under a mild strict feasibility condition, we establish the subsequential convergence of our nonconvex {\color{blue}FW-type} method. Since the feasible region of our generalized LO typically changes from iteration to iteration, our convergence analysis is {\it completely different} from those existing works in the literature on FW-type methods that deal with {\em fixed} feasible regions among subproblems. Finally, motivated by the away steps for accelerating FW-type {\color{blue}methods} for convex problems, we further design an {\it away-step oracle} to supplement our nonconvex {\color{blue}FW-type} method, and establish subsequential convergence of this variant. Numerical results on {\color{blue}the matrix completion problem are presented to demonstrate the efficiency of the proposed FW-type method and its away-step variant via the MovieLens10M dataset.}

3: 

4: 		\keywords{Nonconvex constraint sets \and Frank-Wolfe variants \and generalized linear-optimization oracles \and away-step oracles}

5: 		% \PACS{PACS code1 \and PACS code2 \and more}

6: 		%		\subclass{MSC code1 \and MSC code2 \and more}

7: 	\end{abstract}

8: