1: \begin{abstract}

2: 	In this paper we study the convex problem of optimizing the sum of a smooth function and a compactly supported non-smooth term with a specific separable form. We analyze the block version of the generalized conditional gradient

3: 	method when the blocks are chosen in a cyclic order. A global sublinear rate of

4: 	convergence is established for two different stepsize strategies commonly used in this

5: 	class of methods. Numerical comparisons of the proposed method to both the classical conditional

6: 	gradient algorithm and its random block version demonstrate the

7: 	effectiveness of the cyclic block update rule.

8: \end{abstract}

9: