README.md

CSDP

PackageEvaluator	Build Status

Julia wrapper to CSDP semidefinite programming solver.

The original algorithm is described by B. Borchers. CSDP, A C Library for Semidefinite Programming. Optimization Methods and Software 11(1):613-623, 1999. DOI 10.1080/10556789908805765. Preprint.

Installing CSDP

First, make sure that you have a compiler available and that LAPACK and BLAS are installed. On Ubuntu, simply do

$ sudo apt-get install build-essential liblapack-dev libopenblas-dev

Then, install CSDP using

julia> Pkg.add("CSDP")

To use CSDP with JuMP, do

using JuMP
m = Model(solver=CSDPSolver())

CSDP problem representation

The primal is represented internally by CSDP as follows:

max ⟨C, X⟩
      A(X) = a
         X ⪰ 0

where A(X) = [⟨A_1, X⟩, ..., ⟨A_m, X⟩]. The corresponding dual is:

min ⟨a, y⟩
     A'(y) - C = Z
             Z ⪰ 0

where A'(y) = y_1A_1 + ... + y_mA_m

Termination criteria

CSDP will terminate successfully (or partially) in the following cases:

• If CSDP finds y and Z ⪰ 0 such that -⟨a, y⟩ / ||A'(y) - Z||_F > pinftol, it returns 1 with y such that ⟨a, y⟩ = -1.
• If CSDP finds X ⪰ 0 such that ⟨C, X⟩ / ||A(X)||_2 > dinftol, it returns 2 with X such that ⟨C, X⟩ = 1.
• If CSDP finds X, Z ⪰ 0 such that the following 3 tolerances are satisfied:
  • primal feasibility tolerance: ||A(x) - a||_2 / (1 + ||a||_2) < axtol
  • dual feasibility tolerance: ||A'(y) - C - Z||_F / (1 + ||C||_F) < atytol
  • relative duality gap tolerance: gap / (1 + |⟨a, y⟩| + |⟨C, X⟩|) < objtol
    • objective duality gap: If usexygap is 0, gap = ⟨a, y⟩ - ⟨C, X⟩
    • XY duality gap: If usexygap is 1, gap = ⟨Z, X⟩ then it returns 0.
• If CSDP finds X, Z ⪰ 0 such that the following 3 tolerances are satisfied with 1000*axtol, 1000*atytol and 1000*objtol but at least one of them is not satisfied with axtol, atytol and objtol and cannot make progress then it returns 3.

Remark: In theory, for feasible primal and dual solutions, ⟨a, y⟩ - ⟨C, X⟩ = ⟨Z, X⟩ so the objective and XY duality gap should be equivalent. However, in practice, there are sometimes solution which satisfy primal and dual feasibility tolerances but have objective duality gap which are not close to XY duality gap. In some cases, the objective duality gap may even become negative (hence the tweakgap option). This is the reason usexygap is 1 by default.

Remark: CSDP considers that X ⪰ 0 (resp. Z ⪰ 0) is satisfied when the Cholesky factorizations can be computed. In practice, this is somewhat more conservative than simply requiring all eigenvalues to be nonnegative.

Status

The table below shows how the different CSDP status are converted to MathProgBase status.

CSDP code	State	Description	MathProgBase status
0	Success	SDP solved	Optimal
1	Success	The problem is primal infeasible, and we have a certificate	Infeasible
2	Success	The problem is dual infeasible, and we have a certificate	Unbounded
3	Partial Success	A solution has been found, but full accuracy was not achieved	Suboptimal
4	Failure	Maximum iterations reached	UserLimit
5	Failure	Stuck at edge of primal feasibility	Error
6	Failure	Stuck at edge of dual infeasibility	Error
7	Failure	Lack of progress	Error
8	Failure	X, Z, or O was singular	Error
9	Failure	Detected NaN or Inf values	Error

If the printlevel option is at least 1, the following will be printed:

• If the return code is 1, CSDP will print ⟨a, y⟩ and ||A'(y) - Z||_F
• if the return code is 2, CSDP will print ⟨C, X⟩ and ||A(X)||_F
• otherwise, CSDP will print
  • the primal/dual objective value,
  • the relative primal/dual infeasibility,
  • the objective duality gap ⟨a, y⟩ - ⟨C, X⟩ and objective relative duality gap (⟨a, y⟩ - ⟨C, X⟩) / (1 + |⟨a, y⟩| + |⟨C, X⟩|),
  • the XY duality gap ⟨Z, X⟩ and XY relative duality gap ⟨Z, X⟩ / (1 + |⟨a, y⟩| + |⟨C, X⟩|)
  • and the DIMACS error measures.

Options

The CSDP options are listed in the table below. Their value can be specified in the constructor of the CSDP solver, e.g. CSDPSolver(axtol=1e-7, printlevel=0).

Name		Default Value
axtol	Tolerance for primal feasibility	1.0e-8
atytol	Tolerance for dual feasibility	1.0e-8
objtol	Tolerance for relative duality gap	1.0e-8
pinftol	Tolerance for determining primal infeasibility	1.0e8
dinftol	Tolerance for determining dual infeasibility	1.0e8
maxiter	Limit for the total number of iterations	100
minstepfrac	The minstepfrac and maxstepfrac parameters determine how close to the edge of the feasible region CSDP will step	0.90
maxstepfrac	The minstepfrac and maxstepfrac parameters determine how close to the edge of the feasible region CSDP will step	0.97
minstepp	If the primal step is shorter than minstepp then CSDP declares a line search failure	1.0e-8
minstepd	If the primal step is shorter than minstepp then CSDP declares a line search failure	1.0e-8
usexzgap	If usexzgap is 0 then CSDP will use the objective duality gap d - p instead of the XY duality gap ⟨Z, X⟩	1
tweakgap	If tweakgap is set to 1, and usexzgap is set to 0, then CSDP will attempt to "fix" negative duality gaps	0
affine	If affine is set to 1, then CSDP will take only primal-dual affine steps and not make use of the barrier term. This can be useful for some problems that do not have feasible solutions that are strictly in the interior of the cone of semidefinite matrices	0
perturbobj	The perturbobj parameter determines whether the objective function will be perturbed to help deal with problems that have unbounded optimal solution sets. If perturbobj is 0, then the objective will not be perturbed. If perturbobj is 1, then the objective function will be perturbed by a default amount. Larger values of perturbobj (e.g. 100) increase the size of the perturbation. This can be helpful in solving some difficult problems.	1
fastmode	The fastmode parameter determines whether or not CSDP will skip certain time consuming operations that slightly improve the accuracy of the solutions. If fastmode is set to 1, then CSDP may be somewhat faster, but also somewhat less accurate	0
printlevel	The printlevel parameter determines how much debugging information is output. Use a printlevel of 0 for no output and a printlevel of 1 for normal output. Higher values of printlevel will generate more debugging output	1

Getting the CSDP Library

The original make-file build system only provides a static library. The quite old (September 2010) pycsdp interface by Benjamin Kern circumvents the problem by writing some C++ code to which the static library is linked. The Sage module by @mghasemi is a Cython interface; I don't know how the libcsdp is installed or whether they assume that it is already available on the system.

That is why this package tries to parse the makefile and compiles it itself on Unix systems (so gcc is needed).

For Windows, a pre-compiled DLL is downloaded (unless you configure the build.jl differently).

Next Steps (TODOs)

• Maybe port libcsdp to use 64bit Lapack, aka replace “some ints” by long int (the variables used in a Lapack call). Started in brach julias_openblas64
• Maybe think about an own array type to circumvent the 1-index problems in libcsdp.
• Map Julia's sparse arrays to sparsematrixblock.
• Upload libcsdp.dll for Windows via Appveyor deployment as described at JuliaCon. Currently we use a separate repository.