forum.alglib.net

1. the third parameter supplied to this function 'D' defines the nodal function type.
'0' = constant model (BAD model)
'1' = linear interpolation (OK model)
'2' = quadratic interpolation (BETTER model)

I recently learned at ACM 'TOMS' library, that 'TOMS 790' is an improved BEST 'cubic interpolation' model (preserving C2 continuity) for 2D space dimension problems.
I was wondering if there are any plans to add a type '3' = cubic interpolation type to alglib for this shepards function?
Unfortunately the TOMS library is coded in Fortran, but perhaps its just one small part of IDWBuildModifiedShepard that would need to be added.

2. i am curious to know if such a shepards 'data set' has been calculated having what i'll call 'distinct boundaries', so that within some domain there is regularly spaced nodes, say in the case of a 2D case, such as to interpolate pixel locations of a 2D image per se...
i can understand that the interpolated data 'inside' of this region should give 'proper' answers.
i am curious to know in theory, what happens at the 'edge' of this boundary?
inside of the boundary is valid data, outside of the boundary is 'nothing'.
so a point being interpolated on this 'edge' (between valid data, and nothing) - does shepards simply assume a value of 0 outside of the valid data boundary?
and if so, the interpolated value will be somewhere between the last 'valid data' value and 0?

thanks for explaining.

greg

I recently learned at ACM 'TOMS' library, that 'TOMS 790' is an improved BEST 'cubic interpolation' model (preserving C2 continuity) for 2D space dimension problems.
I was wondering if there are any plans to add a type '3' = cubic interpolation type to alglib for this shepards function?

_________________
.