We propose a Mixed Integer Partial Differential Equation Constrained Optimization (MIPDECO) formulation of the topological optimization for electromagnetic cloaking. Our formulation introduces binary variables to indicate the presence/absence of metamaterial at a given location within the cloaking device. The cloaking device is discretized on a 20×20 and 40×40 domain of squared material locations, and has two to the four-hundred,  two to the sixteen-hundred and  possible configurations respectfully. Sampling methods are useless in creating effective cloaking devices. Instead, we solve the problem by first relaxing the binary variables to be fractional, solving the resulting continuous PDE-constrained optimization problem with a log-barrier approach and an additional regularization term. We then round the relaxed solution for the cloaking device back to 0,1 using a rounding heuristic. Finally we iteratively improve the rounded solution with a 0-1 steepest decent approach. Our method generates new devices that result in better invisibility of regions.