Fast Least Square Approximation in Shift-Invariant Function Spaces

This work addresses the problem of designing fast least square approximation algorithms for shift-invariant function spaces. The proposed scheme relies exclusively on the decomposition property of linear least square approximation problems [2,3], and the shift-invariance of the underlying function space. After showing the potential of reducing a least-square approximation sub-problems of length N, we suggest a recursive divide-and-conquer procedure [1] to minimize the overall computational cost.