SDPLib Importer

This helper function reads the problem data specified in SDPA sparse file format, transforms and saves the problem data into the easier to handle Julia Data Format (JLD). The benchmark problems are provided in the SDPLib 1.2 (a library of semidefinite programming test problems) by Brian Borchers. They are availabe for download: http://infohost.nmt.edu/~borchers/sdplib.html.

Prerequisites

The code is written for Julia v1.0
You need the FileIO and JLD2 package to handle the Julia Data Format (JLD2). To get it simply type add ("JLD2").

Installing / Using

Clone the repository to your local machine.
Add the original SDPLib 1.2 problem files to a folder /sdplib (or edit the path inside SDPLib_Importer.jl).
Run the SDPLib_Importer.jl file. (This will create .jld2 files inside the /sdplib/ folder.)

Testing

The following is a short test script that loads a converted problem in .jld2 format, solves the problem using JuMP and the COSMO.jl solver:

using FileIO, COSMO, SparseArrays, LinearAlgebra, Test, JuMP


data = load("./sdplib/maxG11.jld2");
F = data["F"]
c = data["c"]
m = data["m"]
n = data["n"]
obj_true = data["optVal"]

# Describe primal problem using JuMP and solve with COSMO Solver
model = JuMP.Model(with_optimizer(COSMO.Optimizer, verbose = true, decompose = true));
@variable(model, x[1:m]);
@objective(model, Min, c' * x);
@constraint(model, con1,  Symmetric(-Matrix(F[1]) + sum(Matrix(F[k + 1]) .* x[k] for k in 1:m))  in JuMP.PSDCone());
JuMP.optimize!(model);

Authors

Michael Garstka

License

This project is licensed under the Apache License - see the LICENSE.md file for details.

Questions / Bugs

Please use the Issue Tracker to report bugs or to ask questions.