Merge pull request #147 from DanielVandH/gauss_quad

Add Gauss-Legendre quadrature

Author: ChrisRackauckas

Project.toml

@@ -1,7 +1,7 @@
 name = "Integrals"
 uuid = "de52edbc-65ea-441a-8357-d3a637375a31"
 authors = ["Chris Rackauckas <[email protected]>"]
-version = "3.6.0"
+version = "3.7.0"
 
 [deps]
 CommonSolve = "38540f10-b2f7-11e9-35d8-d573e4eb0ff2"
@@ -26,10 +26,12 @@ Requires = "1"
 SciMLBase = "1.70"
 Zygote = "0.4.22, 0.5, 0.6"
 julia = "1.6"
+FastGaussQuadrature = "0.5"
 
 [extensions]
 IntegralsForwardDiffExt = "ForwardDiff"
 IntegralsZygoteExt = ["Zygote", "ChainRulesCore"]
+IntegralsFastGaussQuadratureExt = "FastGaussQuadrature"
 
 [extras]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
@@ -42,11 +44,13 @@ SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
+FastGaussQuadrature = "442a2c76-b920-505d-bb47-c5924d526838"
 
 [targets]
-test = ["SciMLSensitivity", "StaticArrays", "FiniteDiff", "Pkg", "SafeTestsets", "Test", "Distributions", "ForwardDiff", "Zygote", "ChainRulesCore"]
+test = ["SciMLSensitivity", "StaticArrays", "FiniteDiff", "Pkg", "SafeTestsets", "Test", "Distributions", "ForwardDiff", "Zygote", "ChainRulesCore", "FastGaussQuadrature"]
 
 [weakdeps]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
+FastGaussQuadrature = "442a2c76-b920-505d-bb47-c5924d526838"

docs/src/solvers/IntegralSolvers.md

@@ -11,9 +11,11 @@ The following algorithms are available:
   - `CubaSUAVE`: SUAVE from Cuba.jl. Requires `using IntegralsCuba`.
   - `CubaDivonne`: Divonne from Cuba.jl. Requires `using IntegralsCuba`.
   - `CubaCuhre`: Cuhre from Cuba.jl. Requires `using IntegralsCuba`.
+  - `GaussLegendre`: Uses Gauss-Legendre quadrature with nodes and weights from FastGaussQuadrature.jl.
 
 ```@docs
 QuadGKJL
 HCubatureJL
 VEGAS
+GaussLegendre
 ```

ext/IntegralsFastGaussQuadratureExt.jl

@@ -0,0 +1,51 @@
+module IntegralsFastGaussQuadratureExt
+using Integrals
+if isdefined(Base, :get_extension)
+    import FastGaussQuadrature
+    import FastGaussQuadrature: gausslegendre
+    # and eventually gausschebyshev, etc.
+else
+    import ..FastGaussQuadrature
+    import ..FastGaussQuadrature: gausslegendre
+end
+using LinearAlgebra
+
+Integrals.gausslegendre(n) = FastGaussQuadrature.gausslegendre(n)
+
+function gauss_legendre(f, p, lb, ub, nodes, weights)
+    scale = (ub - lb) / 2
+    shift = (lb + ub) / 2
+    I = dot(weights, @. f(scale * nodes + shift, $Ref(p)))
+    return scale * I
+end
+function composite_gauss_legendre(f, p, lb, ub, nodes, weights, subintervals)
+    h = (ub - lb) / subintervals
+    I = zero(h)
+    for i in 1:subintervals
+        _lb = lb + (i - 1) * h
+        _ub = _lb + h
+        I += gauss_legendre(f, p, _lb, _ub, nodes, weights)
+    end
+    return I
+end
+
+function Integrals.__solvebp_call(prob::IntegralProblem, alg::Integrals.GaussLegendre{C},
+                                  sensealg, lb, ub, p;
+                                  reltol = nothing, abstol = nothing,
+                                  maxiters = nothing) where {C}
+    if isinplace(prob) || lb isa AbstractArray || ub isa AbstractArray
+        error("GaussLegendre only accepts one-dimensional quadrature problems.")
+    end
+    @assert prob.batch == 0
+    @assert prob.nout == 1
+    if C
+        val = composite_gauss_legendre(prob.f, prob.p, lb, ub,
+                                       alg.nodes, alg.weights, alg.subintervals)
+    else
+        val = gauss_legendre(prob.f, prob.p, lb, ub,
+                             alg.nodes, alg.weights)
+    end
+    err = nothing
+    SciMLBase.build_solution(prob, alg, val, err, retcode = ReturnCode.Success)
+end
+end

src/Integrals.jl

@@ -108,15 +108,16 @@ function SciMLBase.solve(prob::IntegralProblem,
     __solvebp(prob, alg, sensealg, prob.lb, prob.ub, prob.p; kwargs...)
 end
 # Throw error if alg is not provided, as defaults are not implemented.
-SciMLBase.solve(::IntegralProblem) = throw(ArgumentError("""
-    No integration algorithm `alg` was supplied as the second positional argument.
-    Reccomended integration algorithms are:
-    For scalar functions: QuadGKJL()
-    For ≤ 8 dimensional vector functions: HCubatureJL()
-    For > 8 dimensional vector functions: MonteCarloIntegration.vegas(f, st, en, kwargs...)
-    See the docstrings of the different algorithms for more detail.
-    """
-))
+function SciMLBase.solve(::IntegralProblem)
+    throw(ArgumentError("""
+No integration algorithm `alg` was supplied as the second positional argument.
+Reccomended integration algorithms are:
+For scalar functions: QuadGKJL()
+For ≤ 8 dimensional vector functions: HCubatureJL()
+For > 8 dimensional vector functions: MonteCarloIntegration.vegas(f, st, en, kwargs...)
+See the docstrings of the different algorithms for more detail.
+"""))
+end
 
 # Give a layer to intercept with AD
 __solvebp(args...; kwargs...) = __solvebp_call(args...; kwargs...)
@@ -188,5 +189,5 @@ function __solvebp_call(prob::IntegralProblem, alg::VEGAS, sensealg, lb, ub, p;
     SciMLBase.build_solution(prob, alg, val, err, chi = chi, retcode = ReturnCode.Success)
 end
 
-export QuadGKJL, HCubatureJL, VEGAS
+export QuadGKJL, HCubatureJL, VEGAS, GaussLegendre
 end # module

src/algorithms.jl

@@ -82,3 +82,43 @@ struct VEGAS <: SciMLBase.AbstractIntegralAlgorithm
     debug::Bool
 end
 VEGAS(; nbins = 100, ncalls = 1000, debug = false) = VEGAS(nbins, ncalls, debug)
+
+"""
+    GaussLegendre{C, N, W}
+
+Struct for evaluating an integral via (composite) Gauss-Legendre quadrature.
+The field `C` will be `true` if `subintervals > 1`, and `false` otherwise.
+
+The fields `nodes::N` and `weights::W` are defined by 
+`nodes, weights = gausslegendre(n)` for a given number of nodes `n`. 
+
+The field `subintervals::Int64 = 1` (with default value `1`) defines the 
+number of intervals to partition the original interval of integration 
+`[a, b]` into, splitting it into `[xⱼ, xⱼ₊₁]` for `j = 1,…,subintervals`,
+where `xⱼ = a + (j-1)h` and `h = (b-a)/subintervals`. Gauss-Legendre 
+quadrature is then applied on each subinterval. For example, if 
+`[a, b] = [-1, 1]` and `subintervals = 2`, then Gauss-Legendre 
+quadrature will be applied separately on `[-1, 0]` and `[0, 1]`,
+summing the two results.
+"""
+struct GaussLegendre{C, N, W} <: SciMLBase.AbstractIntegralAlgorithm
+    nodes::N
+    weights::W
+    subintervals::Int64
+    function GaussLegendre(nodes::N, weights::W, subintervals = 1) where {N, W}
+        if subintervals > 1
+            return new{true, N, W}(nodes, weights, subintervals)
+        elseif subintervals == 1
+            return new{false, N, W}(nodes, weights, subintervals)
+        else
+            throw(ArgumentError("Cannot use a nonpositive number of subintervals."))
+        end
+    end
+end
+function gausslegendre end
+function GaussLegendre(; n = 250, subintervals = 1, nodes = nothing, weights = nothing)
+    if isnothing(nodes) || isnothing(weights)
+        nodes, weights = gausslegendre(n)
+    end
+    return GaussLegendre(nodes, weights, subintervals)
+end

src/init.jl

@@ -2,5 +2,6 @@
     function __init__()
         @require ForwardDiff="f6369f11-7733-5829-9624-2563aa707210" begin include("../ext/IntegralsForwardDiffExt.jl") end
         @require Zygote="e88e6eb3-aa80-5325-afca-941959d7151f" begin include("../ext/IntegralsZygoteExt.jl") end
+        @require FastGaussQuadrature="442a2c76-b920-505d-bb47-c5924d526838" begin include("../ext/IntegralsFastGaussQuadratureExt.jl") end
     end
 end

test/gaussian_quadrature_tests.jl

@@ -0,0 +1,99 @@
+using Integrals, Test, FastGaussQuadrature
+
+#=
+f = (x, p) -> x^3 * sin(5x)
+n = 250
+nodes, weights = gausslegendre(n)
+I = gauss_legendre(f, nothing, -1, 1, nodes, weights)
+@test I ≈ 2 / (625) * (69sin(5) - 95cos(5))
+I = Integrals.composite_gauss_legendre(f, nothing, -1, 1, nodes, weights, 2)
+@test I ≈ 2 / (625) * (69sin(5) - 95cos(5))
+
+f = (x, p) -> (x + p) * abs(x)
+n = 100
+nodes, weights = gausslegendre(n)
+I = Integrals.gauss_legendre(f, 0.0, -2, 2, nodes, weights)
+Ic = Integrals.composite_gauss_legendre(f, 6, -2, 2, nodes, weights, 5)
+@inferred Integrals.gauss_legendre(f, 0.0, -2, 2, nodes, weights)
+@inferred Integrals.composite_gauss_legendre(f, 6, -2, 2, nodes, weights, 5)
+@test I≈0.0 atol=1e-6
+@test Ic≈24 rtol=1e-4
+=#
+
+alg = GaussLegendre()
+n = 250
+nd, wt = gausslegendre(n)
+@test alg.nodes == nd
+@test alg.weights == wt
+@test alg.subintervals == 1
+alg = GaussLegendre(n = 125, subintervals = 3)
+n = 125
+nd, wt = gausslegendre(n)
+@test alg.nodes == nd
+@test alg.weights == wt
+@test alg.subintervals == 3
+@test typeof(alg).parameters[1]
+nd, wt = gausslegendre(275)
+alg = GaussLegendre(nodes = nd, weights = wt)
+@test !typeof(alg).parameters[1]
+@test alg.nodes == nd
+@test alg.weights == wt
+@test alg.subintervals == 1
+alg = GaussLegendre(nodes = nd, weights = wt, subintervals = 20)
+@test typeof(alg).parameters[1]
+@test alg.nodes == nd
+@test alg.weights == wt
+@test alg.subintervals == 20
+
+f = (x, p) -> 5x + sin(x) - p * exp(x)
+prob = IntegralProblem(f, -5, 3, 3.3)
+alg = GaussLegendre()
+sol = solve(prob, alg)
+@test isnothing(sol.chi)
+@test sol.alg === alg
+@test sol.prob === prob
+@test isnothing(sol.resid)
+@test SciMLBase.successful_retcode(sol)
+@test sol.u ≈ -exp(3) * 3.3 + 3.3 / exp(5) - 40 + cos(5) - cos(3)
+alg = GaussLegendre(subintervals = 7)
+sol = solve(prob, alg)
+@test sol.u ≈ -exp(3) * 3.3 + 3.3 / exp(5) - 40 + cos(5) - cos(3)
+
+f = (x, p) -> exp(-x^2) 
+prob = IntegralProblem(f, 0.0, Inf)
+alg = GaussLegendre()
+sol = solve(prob, alg)
+@test sol.u ≈ sqrt(π)/2
+alg = GaussLegendre(subintervals=1)
+@test sol.u ≈ sqrt(π)/2
+alg = GaussLegendre(subintervals=17)
+@test sol.u ≈ sqrt(π)/2
+
+prob = IntegralProblem(f, -Inf, Inf)
+alg = GaussLegendre()
+sol = solve(prob, alg)
+@test sol.u ≈ sqrt(π)
+alg = GaussLegendre(subintervals=1)
+@test sol.u ≈ sqrt(π)
+alg = GaussLegendre(subintervals=17)
+@test sol.u ≈ sqrt(π)
+
+prob = IntegralProblem(f, -Inf, 0.0)
+alg = GaussLegendre()
+sol = solve(prob, alg)
+@test sol.u ≈ sqrt(π)/2
+alg = GaussLegendre(subintervals=1)
+@test sol.u ≈ sqrt(π)/2
+alg = GaussLegendre(subintervals=17)
+@test sol.u ≈ sqrt(π)/2
+
+# Make sure broadcasting correctly handles the argument p 
+f = (x, p) -> 1 + x + x^p[1] - cos(x*p[2]) + exp(x)*p[3] 
+p = [0.3, 1.3, -0.5]
+prob = IntegralProblem(f, 2, 6.3, p)
+alg = GaussLegendre()
+sol = solve(prob, alg)
+@test sol.u ≈ -240.25235266303063249920743158729
+alg = GaussLegendre(n = 500, subintervals = 17)
+sol = solve(prob, alg)
+@test sol.u ≈ -240.25235266303063249920743158729

test/runtests.jl

@@ -13,3 +13,4 @@ dev_subpkg("IntegralsCubature")
 @time @safetestset "Interface Tests" begin include("interface_tests.jl") end
 @time @safetestset "Derivative Tests" begin include("derivative_tests.jl") end
 @time @safetestset "Infinite Integral Tests" begin include("inf_integral_tests.jl") end
+@time @safetestset "Gaussian Quadrature Tests" begin include("gaussian_quadrature_tests.jl") end