Core routines

build_model!(model::JuMP.Model, f_fitness::Function, nx::Int, nh::Int, ng::Int, lx::Vector, ux::Vector, x0::Vector=nothing)

Extend model for NLP problem via memoized fitness function.

Arguments:

• model::JuMP.Model: model to append objective and constraints
• f_fitness::Function: fitness function, returning [f, h, g]
• nx::Int: number of decision variables
• nh::Int: number of equality constraints
• ng::Int: number of inequality constraints
• lx::Vector: lower bounds on decision variables
• ux::Vector: upper bounds on decision variables
• x0::Vector: initial guess
• auto_diff::Bool: whether to use automatic differentiation
• order::Int: order of FiniteDifferences, minimum is 2
• fd_type::String: finite-difference method, "forward", "backward", or "central"

build_model(f_fitness::Function, nx::Int, nh::Int, ng::Int, lx::Vector, ux::Vector, x0::Vector=nothing, fd_type::Function=nothing, order::Int=2)

Build model for NLP problem with memoized fitness function.

Arguments:

• optimizer: optimizer to use with the model
• f_fitness::Function: fitness function, returning [f, h, g]
• nx::Int: number of decision variables
• nh::Int: number of equality constraints
• ng::Int: number of inequality constraints
• lx::Vector: lower bounds on decision variables
• ux::Vector: upper bounds on decision variables
• x0::Vector: initial guess
• auto_diff::Bool: whether to use automatic differentiation
• order::Int: order of FiniteDifferences, minimum is 2
• fd_type::String: finite-difference method, "forward", "backward", or "central"

memoize_fitness(f_fitness::Function, n_outputs::Int)

Memoize fitness function. Because foo_i is auto-differentiated with ForwardDiff, our cache needs to work when x is a Float64 and a ForwardDiff.Dual.

See: https://jump.dev/JuMP.jl/stable/tutorials/nonlinear/tipsandtricks/#Memoization

memoize_fitness_gradient(f_fitness::Function, nfitness::Int, fd_type::Function, order::Int=2)

Create memoized gradient computation with method specified by fd_type - fd_type = "forward" use FiniteDifferences.forward_fdm() - fd_type = "central" use FiniteDifferences.central_fdm() - fd_type = "backward" use FiniteDifferences.backward_fdm()