| |
| | end conditions in univariate spline interpolation (Site not responding. Last check: 2007-10-24) |
 | | Spline interpolation, at knots for even order $k$ and at `half'-knots (i.e., midpoints of knot intervals) for odd order, leaves an even number of degrees of freedom still to be chosen, and usually chosen by prescribing additional conditions at both ends. |
 | | The not-a-knot condition in cubic spline interpolation is easily shown to be correct since, in effect, we are interpolating at $x_1,..., x_n$ by a cubic spline with simple interior knots at $x_3,..., x_{n-2}$, hence the Schoenberg-Whitney conditions are easily seen to hold. |
 | | The explanation is for the simplest case, that of periodic cubic spline interpolation: The data $(x_,y_i)$, $i=1,...,N$ are to be matched by a cubic spline $f$ with period interval $I=[x_1.. |
| www.cs.wisc.edu /~deboor/toast/pages011.html (666 words) |
|