Docs for MOO #823

ParasPuneetSingh · 2024-09-20T16:35:02Z

Checklist

Appropriate tests were added
Any code changes were done in a way that does not break public API
All documentation related to code changes were updated
The new code follows the
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Any new documentation only uses public API

Additional context

Add any other context about the problem here.

docs/src/optimization_packages/blackboxoptim.md

Vaibhavdixit02 · 2024-09-22T04:05:57Z

https://github.com/SciML/Optimization.jl/actions/runs/10977080544/job/30478629355?pr=823 documentation failure seems real, can you take a look?

docs/src/optimization_packages/blackboxoptim.md

Vaibhavdixit02 · 2024-09-23T16:08:29Z

docs/src/optimization_packages/blackboxoptim.md

+end
+mof = MultiObjectiveOptimizationFunction(multi_obj_func_2)
+prob = Optimization.OptimizationProblem(mof, u0; lb = [0.0, 0.0], ub = [2.0, 2.0])
+sol = solve(prob_2, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))


Suggested change

sol = solve(prob_2, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))

sol = solve(prob, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))

Vaibhavdixit02 · 2024-09-23T16:08:56Z

docs/src/optimization_packages/evolutionary.md

+  return [f1, f2]
+end
+initial_guess = [1.0, 1.0]
+function gradient_multi_objective(x, p=nothing)


Since this isn't used please remove it

It is used in the MultiObjectiveOptimizationFunction call below and is passed as the jac arg.

But Evolutionary doesn't need derivatives right?

@ParasPuneetSingh bump

Yeah just checked the testing once again and it is not needed, so will remove it.

Vaibhavdixit02 · 2024-09-23T16:09:50Z

docs/src/optimization_packages/metaheuristics.md

+npartitions = 100
+
+# reference points (Das and Dennis's method)
+weights = gen_ref_dirs(nobjectives, npartitions)


This function is missing here

Vaibhavdixit02 · 2024-09-23T16:10:16Z

@ParasPuneetSingh please take a look at the review above

Vaibhavdixit02 · 2024-09-26T14:05:38Z

Can you rebase on master there was an incorrect compat there that made the docs environment fail here I have updated it

docs/src/optimization_packages/metaheuristics.md

docs/src/optimization_packages/evolutionary.md

ChrisRackauckas · 2024-11-09T19:28:45Z

docs/src/optimization_packages/blackboxoptim.md

 ```
+
+## Multi-objective optimization
+The optimizer for Multi-Objective Optimization is `BBO_borg_moea()`. Your objective function should return a tuple of the objective values and you should indicate the fitness scheme to be (typically) Pareto fitness and specify the number of objectives. Otherwise, the use is similar, here is an example:


Why a tuple? That is not going to scale well.

Yeah corrected that mistake to vector, the struct uses a vector of objective functions.

ChrisRackauckas · 2024-11-09T19:29:05Z

docs/src/optimization_packages/blackboxoptim.md

+end
+mof = MultiObjectiveOptimizationFunction(multi_obj_func)
+prob = Optimization.OptimizationProblem(mof, u0; lb = [0.0, 0.0], ub = [2.0, 2.0])
+sol = solve(prob, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))


NumDimensions, FitnessScheme, those don't match Julia style.

Those should be in the opt.

NumDimensions should be computed from the MultiObjectiveOptimizationFunction, FitnessScheme should be in the opt.

ChrisRackauckas · 2024-11-09T19:29:41Z

docs/src/optimization_packages/evolutionary.md

+function func(x, p=nothing)::Vector{Float64}
+  f1 = (1.0 - x[1])^2 + 100.0 * (x[2] - x[1]^2)^2  # Rosenbrock function
+  f2 = -20.0 * exp(-0.2 * sqrt(0.5 * (x[1]^2 + x[2]^2))) - exp(0.5 * (cos(2π * x[1]) + cos(2π * x[2]))) + exp(1) + 20.0  # Ackley function
+  return [f1, f2]


it's an array now? Is there an in-place form?

ChrisRackauckas · 2024-11-09T19:31:00Z

docs/src/optimization_packages/metaheuristics.md

+    # In this example, we have no constraints
+    gx = [0.0]  # Inequality constraints (not used)
+    hx = [0.0]  # Equality constraints (not used)
+    return [f1, f2], gx, hx


This is now a third API?

ChrisRackauckas · 2024-11-10T04:22:48Z

docs/src/optimization_packages/blackboxoptim.md

+function multi_obj_func(x, p)
+    f1 = (1.0 - x[1])^2 + 100.0 * (x[2] - x[1]^2)^2  # Rosenbrock function
+    f2 = -20.0 * exp(-0.2 * sqrt(0.5 * (x[1]^2 + x[2]^2))) - exp(0.5 * (cos(2π * x[1]) + cos(2π * x[2]))) + exp(1) + 20.0  # Ackley function
+    return (f1, f2)


you didn't actually change it though?

Vaibhavdixit02 · 2025-01-16T13:49:10Z

@ParasPuneetSingh please finish this up.

ParasPuneetSingh · 2025-05-16T07:15:06Z

@ChrisRackauckas please review this at the earliest.

ChrisRackauckas · 2025-05-16T15:00:39Z

docs/src/optimization_packages/blackboxoptim.md

+    f2 = -20.0 * exp(-0.2 * sqrt(0.5 * (x[1]^2 + x[2]^2))) - exp(0.5 * (cos(2π * x[1]) + cos(2π * x[2]))) + exp(1) + 20.0  # Ackley function
+    return (f1, f2)
+end
+mof = MultiObjectiveOptimizationFunction(multi_obj_func)


Suggested change

mof = MultiObjectiveOptimizationFunction(multi_obj_func)

mof = MultiObjectiveOptimizationFunction(multi_obj_func, obj_prototype = zeros(2))

ChrisRackauckas · 2025-05-16T15:01:01Z

docs/src/optimization_packages/blackboxoptim.md

+function multi_obj_func(x, p)
+    f1 = (1.0 - x[1])^2 + 100.0 * (x[2] - x[1]^2)^2  # Rosenbrock function
+    f2 = -20.0 * exp(-0.2 * sqrt(0.5 * (x[1]^2 + x[2]^2))) - exp(0.5 * (cos(2π * x[1]) + cos(2π * x[2]))) + exp(1) + 20.0  # Ackley function
+    return (f1, f2)


Show in-place form too?

ChrisRackauckas · 2025-05-16T15:14:12Z

This is not ready. There's a few steps here. First of all, we document the OptimizationFunction interface https://github.com/SciML/SciMLBase.jl/blob/master/src/scimlfunctions.jl#L1913-L2030 . Right now, MultiObjectiveOptimizationFunction's interface is not documented. Let's start there.

https://github.com/SciML/SciMLBase.jl/blob/master/src/scimlfunctions.jl#L2063-L2066

Write a similar docstring that describes the interface on MultiObjectiveOptimizationFunction. The closest interface should be the NonlinearLeastSquares, since NonlinearLeastSquares is a multi-objective optimization that just has a canonical coalesce via coalesce(cost,u,p) = norm(cost,2).

So okay, what should this thing be? It should be as follows. You have an out of place form cost = f(u,p) and an in-place form f!(cost,u,p). You'll have to have a required cost_prototype which is the vector to use for the in-place cost. This means that length(cost) is equal to the number of objectives. isinplace should then be counting for in-place when you have 3 values, thus matching the same setup as NonlinearFunction. You then need to describe the argument structures expected for each of the derivative definitions, which is more or less matching that of NonlinearLeastSquares. And there should be an optional coalesce(cost,u,p) = norm(cost,2) which when provided allows the MOO problem to be solved with a scalar optimization problem under the definition of coalesce(f(u,p).

Once you have that down, we merge, and then let's use that defined interface. You'll need to MultiobjectiveOptimizationFunction(f; cost_prototype = zeros(2)) some definitions, and then the functions here should not require the user pass in the dimension, since that can be known from the function definition itself.

Vaibhavdixit02 force-pushed the master branch from 3307b06 to 799f7de Compare September 20, 2024 21:51

Vaibhavdixit02 reviewed Sep 22, 2024

View reviewed changes

docs/src/optimization_packages/blackboxoptim.md Outdated Show resolved Hide resolved

Vaibhavdixit02 reviewed Sep 22, 2024

View reviewed changes

docs/src/optimization_packages/blackboxoptim.md Outdated Show resolved Hide resolved

Vaibhavdixit02 reviewed Sep 23, 2024

View reviewed changes

Vaibhavdixit02 closed this Sep 26, 2024

Vaibhavdixit02 reopened this Sep 26, 2024

Vaibhavdixit02 reviewed Sep 27, 2024

View reviewed changes

docs/src/optimization_packages/metaheuristics.md Outdated Show resolved Hide resolved

Vaibhavdixit02 closed this Oct 4, 2024

Vaibhavdixit02 reopened this Oct 4, 2024

Vaibhavdixit02 closed this Oct 26, 2024

Vaibhavdixit02 reopened this Oct 26, 2024

Vaibhavdixit02 reviewed Oct 26, 2024

View reviewed changes

docs/src/optimization_packages/evolutionary.md Outdated Show resolved Hide resolved

Vaibhavdixit02 closed this Oct 28, 2024

Vaibhavdixit02 reopened this Oct 28, 2024

ChrisRackauckas reviewed Nov 9, 2024

View reviewed changes

ChrisRackauckas reviewed Nov 10, 2024

View reviewed changes

ChrisRackauckas reviewed May 16, 2025

View reviewed changes

ParasPuneetSingh closed this May 22, 2025

ParasPuneetSingh force-pushed the master branch from dc2ad91 to b6ce64c Compare May 22, 2025 07:09

	sol = solve(prob_2, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))
	sol = solve(prob, opt, NumDimensions=2, FitnessScheme=ParetoFitnessScheme{2}(is_minimizing=true))

	mof = MultiObjectiveOptimizationFunction(multi_obj_func)
	mof = MultiObjectiveOptimizationFunction(multi_obj_func, obj_prototype = zeros(2))

Uh oh!

Docs for MOO #823

Docs for MOO #823

Uh oh!

Conversation

ParasPuneetSingh commented Sep 20, 2024

Checklist

Additional context

Uh oh!

Uh oh!

Vaibhavdixit02 commented Sep 22, 2024

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Vaibhavdixit02 commented Sep 23, 2024

Uh oh!

Vaibhavdixit02 commented Sep 26, 2024

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Vaibhavdixit02 commented Jan 16, 2025

Uh oh!

ParasPuneetSingh commented May 16, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ChrisRackauckas commented May 16, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants