1: \begin{abstract}

2: % We consider the problem of optimizing a function that may be nonsmooth and nonconvex, subject to bound constraints. Problems of this type arise in a variety of applications including statistics, optimal control and as sub-problems of nonconvex robust optimization problems. The algorithm modifies the traditional L-BFGS algorithm in two ways: (i) it uses an active-set prediction strategy specifically designed for use in the nonsmooth setting along with a corrective strategy that controls the quality of the active sets and generated steps, and (ii) it uses a variant of the weak Wolfe line search to ensure stability in the algorithm. While we do not prove convergence guarantees for the algorithm, we observed that it performs very well in practice. We describe a Python implementation and present numerical results on several test problems to illustrate the efficacy of our approach. 

3: % \end{abstract}

4: