Merge pull request #627 from SciML/Vaibhavdixit02-patch-3

Create polyopt.md
SciML · Nov 17, 2023 · a7ab9f1 · a7ab9f1
2 parents 24660c5 + c386e7c
commit a7ab9f1
Show file tree

Hide file tree

Showing 3 changed files with 31 additions and 0 deletions.
diff --git a/docs/Project.toml b/docs/Project.toml
@@ -24,6 +24,7 @@ OptimizationNOMAD = "2cab0595-8222-4775-b714-9828e6a9e01b"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OptimizationOptimisers = "42dfb2eb-d2b4-4451-abcd-913932933ac1"
 OptimizationPRIMA = "72f8369c-a2ea-4298-9126-56167ce9cbc2"
+OptimizationPolyalgorithms = "500b13db-7e66-49ce-bda4-eed966be6282"
 OptimizationSpeedMapping = "3d669222-0d7d-4eb9-8a9f-d8528b0d9b91"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
 ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
@@ -55,6 +56,7 @@ OptimizationNOMAD = "0.1"
 OptimizationOptimJL = "0.1"
 OptimizationOptimisers = "0.1"
 OptimizationPRIMA = "0.0.1"
+OptimizationPolyalgorithms = "0.1"
 OptimizationSpeedMapping = "0.1"
 OrdinaryDiffEq = "6"
 ReverseDiff = ">= 1.9.0"

diff --git a/docs/pages.jl b/docs/pages.jl
@@ -31,6 +31,7 @@ pages = ["index.md",
         "Optim.jl" => "optimization_packages/optim.md",
         "Optimisers.jl" => "optimization_packages/optimisers.md",
         "PRIMA.jl" => "optimization_packages/prima.md",
+        "Polyalgorithms.jl" => "optimization_packages/polyopt.md",
         "QuadDIRECT.jl" => "optimization_packages/quaddirect.md",
         "SpeedMapping.jl" => "optimization_packages/speedmapping.md",
     ],

diff --git a/docs/src/optimization_packages/polyopt.md b/docs/src/optimization_packages/polyopt.md
@@ -0,0 +1,28 @@
+# OptimizationPolyalgorithms.jl
+
+ OptimizationPolyalgorithms.jl is a package for collecting polyalgorithms formed by fusing popular optimization solvers of different characteristics.
+
+## Installation: OptimizationPolyalgorithms
+
+To use this package, install the OptimizationPolyalgorithms package:
+
+```julia
+import Pkg;
+Pkg.add("OptimizationPolyalgorithms");
+```
+## Algorithms
+
+Right now we support the following polyalgorithms.
+
+`PolyOpt`: Runs Adam followed by BFGS for an equal number of iterations. This is useful in scientific machine learning use cases, by exploring the loss surface with the stochastic optimizer and converging to the minima faster with BFGS.
+
+```@example polyopt
+using Optimization, OptimizationPolyalgorithms
+rosenbrock(x, p) = (p[1] - x[1])^2 + p[2] * (x[2] - x[1]^2)^2
+x0 = zeros(2)
+_p = [1.0, 100.0]
+
+optprob = OptimizationFunction(rosenbrock, Optimization.AutoForwardDiff())
+prob = OptimizationProblem(optprob, x0, _p)
+sol = Optimization.solve(prob, PolyOpt(), maxiters = 1000)
+```