The l1-l2 minimization with rotation for sparse approximation in uncertainty quantification

This paper proposes a combination of rotational compressive sensing with the l1-l2 minimization to estimate coefficients of generalized polynomial chaos (gPC) used in uncertainty quantification. In particular, we aim to identify a rotation matrix such that the gPC of a set of random variables after the rotation has a sparser representation. However, this rotational approach alters the underlying linear system to be solved, which makes finding the sparse coefficients much more difficult than the case without rotation. We further adopt the l1-l2 minimization that is more suited for such ill-posed problems in compressive sensing (CS) than the classic l1 approach. We conduct extensive experiments on standard gPC problem settings, showing superior performance of the proposed combination of rotation and l1-l2 minimization over the ones without rotation and with rotation but using the l1 minimization.