Sparse Grid Interpolation

Employs Smolyak styled grids for sparse interpolation which helps with the curse of dimensionality.

Employs Smolyak styled grids for polynomial and linear interpolation. The depth of grids is adaptive, and with adp=1 is dimension-adaptive. Several features are employed, so that the user can make sure the pairwise dimensions do not outgrow each other (syme, see in file), coefficients that are not relevant get dropped (purge), a title can be added (title), nonvectorized functions can be called up (non_vec), costly functions are not called up twice for the same points (memoize). Error criteria can be adapted.
For full functionality (to see the progress of a nonvectorized job run in parallel) download parfor_progress. And you need to set non_vec=2 and dsp>0. Also, the function afclean which can also be found in the exchange helps to keep the interpolation functions light.