Vector function estimation using splines (Q1091696)

The optimization problem for estimating vector-valued functions is formulated based on a mixed norm, i.e. a linear combination of fidelity and smoothness. Under several sets of distinct covariance assumptions, it is shown that the optimization objective function can be transformed to linear combinations of scalar objective functions for which the solutions are traditional scalar smoothing splines. The vector valued solution to the optimization problem can thus be written as a vector of linear combinations of scalar smoothing splines. Several applications are discussed.