-
Notifications
You must be signed in to change notification settings - Fork 3
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #13 from jrising/smoothopt
Adds smooth optimization functionality
- Loading branch information
Showing
5 changed files
with
224 additions
and
84 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,89 @@ | ||
| using Interpolations | ||
| import Dierckx | ||
| using Distributions | ||
|
|
||
| using NLopt | ||
| # import OptiMimi.OptimizationProblem, OptiMimi.getdims, OptiMimi.gradfreeobjective | ||
| # using Plots | ||
|
|
||
| function piecewisesolution_single(optprob::OptimizationProblem, initgen::Function, maxiter=Inf, verbose=false; reltol=1e-6) | ||
| len = getdims(optprob.model, optprob.components[1], optprob.names[1])[1] | ||
| numsegs = Int64.([2 .^(0:(floor(log2(len-1))-1)); len-1]) | ||
|
|
||
| last_xx = last_soln = nothing | ||
|
|
||
| for numseg in numsegs | ||
| print(numseg) | ||
| xx = trunc.(Int, range(1, stop=len, length=numseg+1)) | ||
|
|
||
| keeps = [true for ii in 1:length(xx)] | ||
| if numseg > 2 | ||
| # Drop xx where [LinearInterp - CubicInterp] < reltol | ||
| linears = LinearInterpolation(last_xx, last_soln)(xx) | ||
| cubics = Dierckx.evaluate(Dierckx.Spline1D(last_xx, last_soln, k=min(3, length(last_xx) - 1)), xx) | ||
| keeps = abs.(linears .- cubics) ./ ((linears .+ cubics) / 2) .> reltol | ||
| keeps[1] = true | ||
| keeps[length(keeps)] = true | ||
|
|
||
| # Drop knots only if can be linearly guessed | ||
| for ii in 2:numseg | ||
| if xx[ii] ∈ last_xx | ||
| jj = findfirst(last_xx .== xx[ii]) | ||
| guess = LinearInterpolation([last_xx[jj-1], last_xx[jj+1]], [last_soln[jj-1], last_soln[jj+1]])(xx[ii]) | ||
| if abs(guess - last_soln[jj]) / last_soln[jj] > reltol | ||
| keeps[ii] = true | ||
| end | ||
| end | ||
| end | ||
|
|
||
| println(keeps) | ||
| end | ||
|
|
||
| paramtrans = convert(Vector{Function}, [yy -> LinearInterpolation(xx[keeps], yy)(1:len)]) | ||
|
|
||
| opt = Opt(optprob.opt.algorithm, sum(keeps)) | ||
| lower_bounds!(opt, optprob.opt.lower_bounds[xx[keeps]]) | ||
| upper_bounds!(opt, optprob.opt.upper_bounds[xx[keeps]]) | ||
| xtol_rel!(opt, reltol) | ||
|
|
||
| myobjective = gradfreeobjective(optprob.model, optprob.components, optprob.names, optprob.objective, paramtrans) | ||
|
|
||
| max_objective!(opt, myobjective) | ||
|
|
||
| for constraint in optprob.constraints | ||
| let this_constraint = constraint | ||
| function my_constraint(xx::Vector, grad::Vector) | ||
| setparameters(optprob.model, optprob.components, optprob.names, xx, paramtrans) | ||
| run(optprob.model) | ||
| this_constraint(optprob.model) | ||
| end | ||
|
|
||
| inequality_constraint!(opt, my_constraint) | ||
| end | ||
| end | ||
|
|
||
| if sum(keeps) == len - 1 | ||
| subprob = OptimizationProblem(optprob.model, optprob.components, optprob.names, optprob.objective, opt, optprob.constraints, nothing) | ||
| else | ||
| subprob = OptimizationProblem(optprob.model, optprob.components, optprob.names, optprob.objective, opt, optprob.constraints, paramtrans) | ||
| end | ||
|
|
||
| if isnothing(last_xx) | ||
| lin_yy = nothing | ||
| else | ||
| lin_yy = LinearInterpolation(last_xx, last_soln)(xx[keeps]) | ||
| end | ||
| soln = solution(subprob, initgen(sum(keeps), lin_yy)) | ||
|
|
||
| if numseg > 2 | ||
| last_soln = LinearInterpolation(last_xx, last_soln)(xx) | ||
| last_soln[keeps] = soln[2] | ||
| else | ||
| last_soln = soln[2] | ||
| end | ||
| last_xx = xx | ||
| println([last_xx, last_soln]) | ||
| end | ||
|
|
||
| last_soln | ||
| end |
Oops, something went wrong.