A comparison of MARS and SLAM in approximate optimization problems

The problem of numerically minimizing a functional cost typically involves the following two issues: (i) the choice of a class of models to approximate the solution; (ii) the definition of a sampling of the domain where the functional is evaluated. This work presents a comparison of performances given by two efficient techniques for function estimation, namely Multi-Adaptive Regression Spline (MARS) and Semi-Local Approximate Minimization (SLAM) in different functional optimization contexts.