1: \begin{abstract}

2: We analyze a solution method for minimization problems over a space of $\R^d$-valued functions of bounded variation on an interval~$I$. The presented method relies on piecewise constant iterates. In each iteration the algorithm alternates between proposing a new point at which the iterate is allowed to be discontinuous and optimizing the magnitude of its jumps as well as the offset. A sublinear~$\mathcal{O}(1/k)$ convergence rate for the objective function values is obtained in general settings. Under additional structural assumptions on the dual variable this can be improved to a locally linear rate of convergence~$\mathcal{O}(\zeta^k)$ for some~$\zeta <1$. Moreover, in this case, the same rate can be expected for the iterates in~$L^1(I;\R^d)$.

3: \end{abstract}

4: