I am trying to find the Chebychev approximation nodes on a triangle (or tetrahedron) with arbitrary edge lengths.
For example
Suppose I have a $D$-variate function, $f(x_{1},...,x_{D})$, with range in $\mathbb{R}^1$.
The domain of $f$ is a tetrahedron, with constraints as follows:
$x_{1}+x_{2}\leq a_1$
$x_{3}+x_{4}\leq a_2$
$\vdots$
$x_{D-1}+x_{D}\leq a_D$
and
$x_i\geq 0 \;\forall \; i={1,...,D}$
(Alternatively a simpler tetrahedron I'd also be interested in would be $\sum_{i=1}^Dx_{i}<a$)
For example, suppose $D=2$ and $a_{1}=25$, how would I find the Chebychev nodes on this space below?
The chebychev nodes on the full rectangle are:
As you can see, the edges are oversampled to minimize approximation error. If I now just take the domain restrictions, we end up with seemingly too few points along the hypotenuse!
