SimplicialCubature.jl documentation

This package is a port of the R package SimplicialCubature, written by John P. Nolan, and which contains R translations of some Matlab and Fortran code written by Alan Genz.

___

A simplex is a triangle in dimension 2, a tetrahedron in dimension 3. This package provides two main functions: integrateOnSimplex, to integrate an arbitrary function on a simplex, and integratePolynomialOnSimplex, to get the exact value of the integral of a multivariate polynomial on a simplex.

A $n$-dimensional simplex must be given by $n+1$ vectors of length $n$, which represent the simplex vertices. For example, the $3$-dimensional unit simplex is encoded as follows:

S = [[0, 0, 0], [1, 0, 0], [0, 1, 0], [0, 0, 1]]

Or you can get it by running CanonicalSimplex(3).

Suppose you want to integrate the function

\[f(x, y ,z) = x + yz\]

on the unit simplex. To use integrateOnSimplex, you have to define $f$ as a function of a 3-dimensional vector:

function f(x)
  return x[1] + x[2]*x[3]
end

using SimplicialCubature
I = integrateOnSimplex(f, S)

Then the value of the integral is given in I.integral.

Since the function $f$ of this example is polynomial, you can use integratePolynomialOnSimplex:

using SimplicialCubature
using TypedPolynomials

@polyvar x y z
P = x + y*z
integratePolynomialOnSimplex(P, S)

Be careful if your polynomial does not involve one of the variables. For example if $P(x, y, z) = x + y$, you have to encode it as a polynomial depending on $z$: type P = x + y + 0*z.

In addition, on this example where the vertex coordinates of $S$ and the coefficients of $P$ are integer numbers, there is a more clever way to proceed: while integratePolynomialOnSimplex implements an exact proedure, it is not free of (small) numerical errors, but the returned value in this situation will be really exact if you use a polynomial with rational coefficients:

@polyvar x y z
P = 1//1*x + y*z
integratePolynomialOnSimplex(P, S)

Member functions

SimplicialCubature.integrateOnSimplexMethod
integrateOnSimplex(f, S; dim, maxEvals, absError, tol, rule, info, fkwargs...)

Integration of a function over one or more simplices.

Arguments

  • f: function to be integrated; must return a real scalar value or a real vector
  • S: simplex or vector of simplices; a simplex is given by n+1 vectors of dimension n
  • dim: number of components of f
  • maxEvals: maximum number of calls to f
  • absError: requested absolute error
  • tol: requested relative error
  • rule: integration rule, an integer between 1 and 4; a 2*rule+1 degree integration rule is used
  • info: Boolean, whether to print more info
  • fkwargs: keywords arguments of f
source

References

  • A. Genz and R. Cools.

An adaptive numerical cubature algorithm for simplices. ACM Trans. Math. Software 29, 297-308 (2003).

  • Jean B. Lasserre.

Simple formula for the integration of polynomials on a simplex. BIT Numerical Mathematics 61, 523-533 (2021).