Smoothing noisy data with spline functions (Q5902942)

The authors present a procedure with a polynomial spline of degree 2m-1 for smoothing noisy data values observed at n distinct not necessarily equally spaced finite intervals. The procedure which minimizes the trace of the influence matrix associated with the smoothing spline requires just order \(m^ 2n\) operations, including the generalized cross validation, and order mn storage locations. Other methods require order \(n^ 3\) operations, therefore the described procedure is a significant improvement.