https://github.com/yuricst/nlpsaut
Wrapper for constructing NLP with Julia
https://github.com/yuricst/nlpsaut
Last synced: 4 months ago
JSON representation
Wrapper for constructing NLP with Julia
- Host: GitHub
- URL: https://github.com/yuricst/nlpsaut
- Owner: Yuricst
- License: mit
- Created: 2021-09-25T23:37:14.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2024-12-12T04:53:07.000Z (over 1 year ago)
- Last Synced: 2025-04-04T10:16:34.426Z (about 1 year ago)
- Language: Julia
- Homepage: https://yuricst.github.io/NLPSaUT/
- Size: 610 KB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# NLPSaUT
📚[Read the docs here!](https://yuricst.github.io/NLPSaUT/)📚
The `NLPSaUT` module constructs a `JuMP` `model` for a generic nonlinear program (NLP).
The expected use case is solving a differentiable (either analytically or numerically) nonconvex NLP with gradient-based algorithms such as Ipopt or SNOPT.
The user is expected to provide a "fitness function" ([pygmo](https://github.com/esa/pygmo2)-style), which evaluates the objective, equality, and inequality constraints. Below is an example:
```julia
function f_fitness(x::T...) where {T<:Real}
# compute objective
f = x[1]^2 - x[2]
# equality constraints
h = zeros(T, 1)
h = x[1]^3 + x[2] - 2.4
# inequality constraints
g = zeros(T, 2)
g[2] = -0.3x[1] + x[2] - 2 # y <= 0.3x + 2
g[1] = x[1] + x[2] - 5 # y <= -x + 5
return [f; h; g]
end
```
Derivatives of `f_fitness` is taken using `ForwardDiff.jl` (which is the default `JuMP` behavior according to its [docs](https://jump.dev/JuMP.jl/stable/tutorials/nonlinear/operator_ad/#Gradient)); as such, `f_fitness` should be written in a way that is compatiable to `ForwardDiff.jl` ([read here as to why it is `ForwardDiff`, not `ReverseDiff`](https://jump.dev/JuMP.jl/stable/manual/nonlinear/#Automatic-differentiation-2)).
For reference, here's the [JuMP docs page on common mistakes when using `ForwardDiff.jl`](https://jump.dev/JuMP.jl/stable/manual/nonlinear/#Common-mistakes-when-writing-a-user-defined-operator).
The `model` constructed by `NLPSaUT` utilizes `memoization` to economize on the fitness evaluation (see [JuMP Tips and tricks on NLP](https://jump.dev/JuMP.jl/stable/tutorials/nonlinear/tips_and_tricks/)).
## Quick start
1. `git clone` this repository
2. start julia-repl
3. activate & instantiate package (first time)
```julia-repl
pkg> activate .
julia> using Pkg # first time only
julia> Pkg.instantiate() # first time only
```
4. run tests
```julia-repl
(NLPSaUT) pkg> test
```
### Note
- To use with SNOPT, it's probably better to go through [`GAMS.jl`](https://github.com/GAMS-dev/gams.jl?tab=readme-ov-file) rather than to use [`SNOPT7.jl`](https://github.com/snopt/SNOPT7.jl) directly (installing `SNOPT7.jl` currently errors on julia v1.10)
## Examples
For examples, see the `examples` directory.
## TODO notes
- Development
- [ ] Finite difference gradient option
- [ ] Analytical gradient option
- Examples/documentation
- [x] Simple example with Ipopt
- [ ] Simple example with SNOPT via GAMS
- [x] Example with ODEProblem
- [ ] Demonstration of memoization benefits
- [ ] Parallelized fitness function