Discrete least-squares approximations over optimized downward closed polynomial spaces in arbitrary dimension (Q2359687)

At first the authors introduce the necessary notations and some of the previous results achieved in the analysis of discrete least squares on fixed multivariate polynomial spaces. The main results are in connection with the discrete least-squares approximations on optimized polynomial spaces, establishing upper bounds on the number of downward closed or anchored sets of a given cardinality, or on the cardinality of their union. The selection of the optimal polynomial space is based on minimizing the least-squares error among all possible choices of downward closed or anchored sets of a given cardinality. In the last part the alternative algorithms for model selection are presented which is a general discussion on alternate selection strategies with reasonable computational cost.