A fast non-recursive algorithm and a parallel processor architecture for smoothing spline fitting

An operator factorization algorithm is proposed for the solution of the smoothing spline fitting problem, when the data are impaired by noise. The algorithm is decomposed into a number of independent procedures which can be realised concurrently by a corresponding set of processors, running in a parallel arrangement and operating on the whole set of the given data points (nodes set). The processor architecture is described and a complexity analysis is presented in terms of the computational complexity and storage requirements for the algorithm realization.