How to generate a vectors, uniformely distributed into convex polytope? I awared of the solution of reduced problem: uniform sampling on a unit simplex. Also I know how to split (even non-convex) polytope into the simplicial pieces (by means of Delaunay triangulation). Having such splitting I can simply work with particular simplicies, considering its hypervolumes as weights.
But how to deal with just a simplicies? I can't simply deform the unit simplex (internals of which consists of random points) to give the arbitrary simplex only using linear transformations. The uniformity of spatial distribution is violated in such case.
I awared that Dirichlet distribution is connected to the problem. But I do not know the means of how to.
I suspect that there is some functional (maybe linear?) dependency between parametres of Gamma-distributions of the components and dot products of radius-vectors of simplex vertices (Gramian matrix, just a supposition).
EDIT:
I want to specifically note, that only the rejection-step-less algorithms are interested. Or, another words, optimal in sense of minimum of amount of generated uniformely distributed "source" values.
EDIT:
Finally I found the solution.