Merge branch 'master' into as/fix-sii

Project.toml

@@ -1,7 +1,7 @@
 name = "NonlinearSolve"
 uuid = "8913a72c-1f9b-4ce2-8d82-65094dcecaec"
 authors = ["SciML"]
-version = "4.1.0"
+version = "4.2.0"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
@@ -96,9 +96,9 @@ NonlinearSolveSpectralMethods = "1"
 OrdinaryDiffEqTsit5 = "1.1.0"
 PETSc = "0.3"
 Pkg = "1.10"
+PolyesterForwardDiff = "0.1"
 PrecompileTools = "1.2"
 Preferences = "1.4"
-PolyesterForwardDiff = "0.1"
 Random = "1.10"
 ReTestItems = "1.24"
 Reexport = "1.2"

lib/NonlinearSolveFirstOrder/Project.toml

@@ -1,7 +1,7 @@
 name = "NonlinearSolveFirstOrder"
 uuid = "5959db7a-ea39-4486-b5fe-2dd0bf03d60d"
 authors = ["Avik Pal <[email protected]> and contributors"]
-version = "1.0.0"
+version = "1.1.0"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
@@ -13,6 +13,7 @@ FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 LinearSolve = "7ed4a6bd-45f5-4d41-b270-4a48e9bafcae"
+LineSearch = "87fe0de2-c867-4266-b59a-2f0a94fc965b"
 MaybeInplace = "bb5d69b7-63fc-4a16-80bd-7e42200c7bdb"
 NonlinearSolveBase = "be0214bd-f91f-a760-ac4e-3421ce2b2da0"
 PrecompileTools = "aea7be01-6a6a-4083-8856-8a6e6704d82a"

lib/NonlinearSolveFirstOrder/src/NonlinearSolveFirstOrder.jl

@@ -8,6 +8,7 @@ using Setfield: @set!
 using ADTypes: ADTypes
 using ArrayInterface: ArrayInterface
 using LinearAlgebra: LinearAlgebra, Diagonal, dot, diagind
+using LineSearch: BackTracking
 using StaticArraysCore: SArray
 
 using CommonSolve: CommonSolve
@@ -19,7 +20,7 @@ using NonlinearSolveBase: NonlinearSolveBase, AbstractNonlinearSolveAlgorithm,
                           AbstractDampingFunctionCache, AbstractTrustRegionMethod,
                           AbstractTrustRegionMethodCache,
                           Utils, InternalAPI, get_timer_output, @static_timeit,
-                          update_trace!, L2_NORM,
+                          update_trace!, L2_NORM, NonlinearSolvePolyAlgorithm,
                           NewtonDescent, DampedNewtonDescent, GeodesicAcceleration,
                           Dogleg
 using SciMLBase: SciMLBase, AbstractNonlinearProblem, NLStats, ReturnCode,
@@ -36,6 +37,8 @@ include("levenberg_marquardt.jl")
 include("trust_region.jl")
 include("pseudo_transient.jl")
 
+include("poly_algs.jl")
+
 include("solve.jl")
 
 @setup_workload begin
@@ -100,4 +103,7 @@ export RadiusUpdateSchemes
 
 export GeneralizedFirstOrderAlgorithm
 
+# Polyalgorithms
+export RobustMultiNewton
+
 end

lib/NonlinearSolveFirstOrder/src/poly_algs.jl

@@ -0,0 +1,44 @@
+"""
+    RobustMultiNewton(
+        ::Type{T} = Float64;
+        concrete_jac = nothing,
+        linsolve = nothing,
+        autodiff = nothing, vjp_autodiff = nothing, jvp_autodiff = nothing
+    )
+
+A polyalgorithm focused on robustness. It uses a mixture of Newton methods with different
+globalizing techniques (trust region updates, line searches, etc.) in order to find a
+method that is able to adequately solve the minimization problem.
+
+Basically, if this algorithm fails, then "most" good ways of solving your problem fail and
+you may need to think about reformulating the model (either there is an issue with the model,
+or more precision / more stable linear solver choice is required).
+
+### Arguments
+
+  - `T`: The eltype of the initial guess. It is only used to check if some of the algorithms
+    are compatible with the problem type. Defaults to `Float64`.
+"""
+function RobustMultiNewton(
+        ::Type{T} = Float64;
+        concrete_jac = nothing,
+        linsolve = nothing,
+        autodiff = nothing, vjp_autodiff = nothing, jvp_autodiff = nothing
+) where {T}
+    common_kwargs = (; concrete_jac, linsolve, autodiff, vjp_autodiff, jvp_autodiff)
+    if T <: Complex # Let's atleast have something here for complex numbers
+        algs = (
+            NewtonRaphson(; common_kwargs...),
+        )
+    else
+        algs = (
+            TrustRegion(; common_kwargs...),
+            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.Bastin),
+            NewtonRaphson(; common_kwargs...),
+            NewtonRaphson(; common_kwargs..., linesearch = BackTracking()),
+            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.NLsolve),
+            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.Fan)
+        )
+    end
+    return NonlinearSolvePolyAlgorithm(algs)
+end

src/NonlinearSolve.jl

@@ -25,7 +25,7 @@ using StaticArraysCore: StaticArray
 
 # Default Algorithm
 using NonlinearSolveFirstOrder: NewtonRaphson, TrustRegion, LevenbergMarquardt, GaussNewton,
-                                RUS
+                                RUS, RobustMultiNewton
 using NonlinearSolveQuasiNewton: Broyden, Klement
 using SimpleNonlinearSolve: SimpleBroyden, SimpleKlement
 
@@ -125,8 +125,7 @@ end
 @reexport using LinearSolve
 
 # Poly Algorithms
-export NonlinearSolvePolyAlgorithm,
-       RobustMultiNewton, FastShortcutNonlinearPolyalg, FastShortcutNLLSPolyalg
+export NonlinearSolvePolyAlgorithm, FastShortcutNonlinearPolyalg, FastShortcutNLLSPolyalg
 
 # Extension Algorithms
 export LeastSquaresOptimJL, FastLevenbergMarquardtJL, NLsolveJL, NLSolversJL,

src/poly_algs.jl

@@ -1,48 +1,3 @@
-"""
-    RobustMultiNewton(
-        ::Type{T} = Float64;
-        concrete_jac = nothing,
-        linsolve = nothing,
-        autodiff = nothing, vjp_autodiff = nothing, jvp_autodiff = nothing
-    )
-
-A polyalgorithm focused on robustness. It uses a mixture of Newton methods with different
-globalizing techniques (trust region updates, line searches, etc.) in order to find a
-method that is able to adequately solve the minimization problem.
-
-Basically, if this algorithm fails, then "most" good ways of solving your problem fail and
-you may need to think about reformulating the model (either there is an issue with the model,
-or more precision / more stable linear solver choice is required).
-
-### Arguments
-
-  - `T`: The eltype of the initial guess. It is only used to check if some of the algorithms
-    are compatible with the problem type. Defaults to `Float64`.
-"""
-function RobustMultiNewton(
-        ::Type{T} = Float64;
-        concrete_jac = nothing,
-        linsolve = nothing,
-        autodiff = nothing, vjp_autodiff = nothing, jvp_autodiff = nothing
-) where {T}
-    common_kwargs = (; concrete_jac, linsolve, autodiff, vjp_autodiff, jvp_autodiff)
-    if T <: Complex # Let's atleast have something here for complex numbers
-        algs = (
-            NewtonRaphson(; common_kwargs...),
-        )
-    else
-        algs = (
-            TrustRegion(; common_kwargs...),
-            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.Bastin),
-            NewtonRaphson(; common_kwargs...),
-            NewtonRaphson(; common_kwargs..., linesearch = BackTracking()),
-            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.NLsolve),
-            TrustRegion(; common_kwargs..., radius_update_scheme = RUS.Fan)
-        )
-    end
-    return NonlinearSolvePolyAlgorithm(algs)
-end
-
 """
     FastShortcutNonlinearPolyalg(
         ::Type{T} = Float64;