The maximum entropy method on the mean: Applications to linear programming and superresolution

Fabrice Gamboa, Elisabeth Gassiat

Abstract: In this paper, we give two different results. We propose new methods to solve classical optimization problems in linear programming. We also obtain precise quantitative results for the superresolution phenomenon, as observed earlier by practical searchers on specific algorithms. The common background of our work is the generalized moment problem, which is known to be connected with linear programming and superresolution. We describe the Maximum Entropy Method on the Mean that provides solution to the problem and leads to computational criteria to decide the existence of solutions or not.