Proposed method makes a number of simplifying assumptions which convert the EEG/FMRI integration problem into optimization of a convex function, of a form amenable to efficient solution as a very sparse linear programming (LP) problem. The assumptions made in doing this are, surprisingly, in general somewhat more robust than those generally used to cast EEG/FMRI integration as optimization of a non-convex function not amenable to efficient global optimization. This is because the L1 norm used here corresponds to a more robust statistical estimator than the L2 normal generally used For this reason, even though this technique results in a tractable global optimization, it is more robust to non-Gaussian noise and outliers than approaches that make the Gaussian noise assumption [1]. Current poster presents formulation of the problem together with results obtained on artificial data.