SCIP-SDP
2.1.0
|
The following important parameters (with these default values) were added:
relaxing/SDP/freq = 1 | set this to -1 and lp/solvefreq to 1 to solve LP relaxations with eigenvector cuts |
constraints/SDP/threads = 1 | number of threads used for openblas; only available with OpenBLAS/OpenMP and compile options SDPS=sdpa and MKL=true (default for SDPS=sdpa) |
relaxing/SDP/displaystatistics = TRUE | Should statistics about SDP iterations and solver settings/success be printed after quitting SCIP-SDP ? |
relaxing/SDP/slatercheck = 0 | Should the Slater condition for the primal and dual problem be checked ahead of solving each SDP? [0: no, 1: yes and output statistics, 2: yes and print warning for every problem not satisfying primal and dual Slater condition] |
relaxing/SDP/sdpinfo = FALSE | Should output of the SDP-Solver be printed to the console? |
branching/sdpinfobjective/coupledvars = FALSE | If all branching candidates have objective zero, should we use the sum of the absolute objectives of all continuous variables coupled with the candidate through constraints? |
branching/sdpinfobjective/singlecoupledvars = FALSE | If all branching candidates have objective zero, should we use the sum of the absolute objectives of all continuous variables coupled with the candidate through constraints in which no other candidate appears? |
branching/sdpobjective/coupledvars = FALSE | If all branching candidates have objective zero, should we use the sum of the absolute objectives of all continuous variables coupled with the candidate through constraints? |
branching/sdpobjective/singlecoupledvars = FALSE | If all branching candidates have objective zero, should we use the sum of the absolute objectives of all continuous variables coupled with the candidate through constraints in which no other candidate appears? |
display/sdpfastsettings/active = 0 | Should the percentage of SDP-relaxations solved with the fastest setting (SDPA) or the default formulation (DSDP) be displayed in the console? [0: off, 1: auto, 2:on] |
display/sdppenalty/active = 0 | Should the percentage of SDP-relaxations solved using a penalty formulation be displayed in the console? [0: off, 1: auto, 2:on] |
display/sdpunsolved/active = 1 | Should the percentage of SDP-relaxations that could not be solved be displayed in the console? [0: off, 1: auto, 2:on] |
heuristics/sdpfracdiving/freq = -1 | set this to 0 or more to enable a fractional diving heuristic for SDPs |
heuristics/sdprand/freq = 10 | set this to -1 to disable the randomized rounding heuristic |
heuristics/sdprand/generalints = FALSE | Should randomized rounding also be applied if there are general integer variables and not only binary variables ? |
heuristics/sdprand/nrounds = 5 | number of rounding rounds |
propagating/sdp-obbt/freq = -1 | set this to 0 or more to enable optimization-based bound tightening using SDP-relaxations |
propagating/sdp-obbt/propcont = TRUE | Should optimization-based bound tightening be performed for continuous variables ? |
propagating/sdp-obbt/propbin = FALSE | Should optimization-based bound tightening be performed for binary variables ? |
propagating/sdpredcost/freq = 1 | set this to -1 to disable reduced cost fixing for SDPs |
propagating/sdpredcost/forbins = TRUE | Should SDP reduced cost fixing be executed for binary variables? |
propagating/sdpredcost/forintcons = TRUE | Should SDP reduced cost fixing be executed for integer and continuous variables? |
relaxing/SDP/lambdastar = -1 | the parameter lambda star used by SDPA to set the initial point , set this to a negative value to compute the parameter depending on the given problem |
relaxing/SDP/maxpenaltyparam = -1 | the maximum value of the penalty parameter Gamma used for the penalty formulation if the SDP solver didn't converge; set this to a negative value to compute the parameter depending on the given problem |
relaxing/SDP/objlimit = FALSE | Should an objective limit be given to the SDP-Solver? |
relaxing/SDP/penaltyparam = -1 | the starting value of the penalty parameter Gamma used for the penalty formulation if the SDP solver didn't converge; set this to a negative value to compute the parameter depending on the given problem |
relaxing/SDP/resolve = TRUE | Should the relaxation be resolved after bound-tightenings were found during propagation (outside of probing)? |
relaxing/SDP/sdpsolverepsilon = 1e-04 | sets the bound for the duality gap in the SDP-Solver |
relaxing/SDP/sdpsolverfeastol = 1e-06 | feasibility tolerance for the SDP-Solver (should be less or equal to numerics/feastol) |
relaxing/SDP/settingsresetfreq = -1 | frequency for resetting parameters in SDP solver and trying again with fastest settings [-1: never, 0: only at depth settingsresetofs, n: all nodes with depth a multiple of n] |
relaxing/SDP/settingsresetofs = 0 | frequency offset for resetting parameters in SDP solver and trying again with fastest settings |
relaxing/SDP/tightenvb = TRUE | Should Big-Ms in varbound-like constraints be tightened before giving them to the SDP-solver ? |