Spurious minimizers in non uniform Fourier sampling optimization

Abstract A recent trend in the signal/image processing literature is optimization of Fourier sampling schemes for specific datasets signals. In this paper, we explain why choosing optimal non Cartesian patterns a difficult nonconvex problem by bringing to light two issues. The first one existence combinatorial number spurious minimizers generic class second vanishing gradient effect high frequencies. We conclude paper showing how using large can mitigate and illustrate experimentally benefits stochastic algorithms with variable metric.