|Date:||Tue, November 20, 2012|
|Place:||Research I Seminar Room|
Abstract: A common problem arising in many optical measurements is signal reconstruction from the absolute value of Fourier coefficients. In the discrete setting this means to reconstruct a signal from the magnitudes of one-dimensional projections. We consider a more general problem and aim to reconstruct a signal from the norms of higher dimensional subspace components. First, we derive a closed formula for reconstruction. Second, we use random subspaces and semidefinite programming to reduce the number of subspace components.