kalmarek
commented
2 years ago btw. with CSDP I see
CSDP 6.2.0
Iter: 0 Ap: 0.00e+00 Pobj: 0.0000000e+00 Ad: 0.00e+00 Dobj: 0.0000000e+00
Iter: 1 Ap: 7.31e-01 Pobj: -5.9540730e+00 Ad: 8.42e-01 Dobj: 3.0671681e-01
Iter: 2 Ap: 8.58e-01 Pobj: -1.6732383e+01 Ad: 8.59e-01 Dobj: 7.8448175e+00
Iter: 3 Ap: 8.56e-01 Pobj: -6.5657908e+00 Ad: 7.36e-01 Dobj: 4.4153069e+00
Iter: 4 Ap: 7.93e-01 Pobj: -3.0377724e+00 Ad: 8.60e-01 Dobj: 8.3889222e-01
Iter: 5 Ap: 8.47e-01 Pobj: -1.8655047e+00 Ad: 8.10e-01 Dobj: -2.9577552e-01
[ ... ]
Iter: 20 Ap: 5.01e-01 Pobj: -9.3179641e-01 Ad: 4.21e-01 Dobj: -9.3279199e-01
Iter: 21 Ap: 1.00e+00 Pobj: -9.3176281e-01 Ad: 1.00e+00 Dobj: -9.3285541e-01
Iter: 22 Ap: 6.71e-01 Pobj: -9.3198047e-01 Ad: 5.60e-01 Dobj: -9.3288054e-01
Iter: 23 Ap: 1.00e+00 Pobj: -9.3188336e-01 Ad: 1.00e+00 Dobj: -9.3289711e-01
Iter: 24 Ap: 9.42e-01 Pobj: -9.3195825e-01 Ad: 8.76e-01 Dobj: -9.3290304e-01
Iter: 25 Ap: 1.00e+00 Pobj: -9.3197421e-01 Ad: 1.00e+00 Dobj: -9.3290342e-01
Iter: 26 Ap: 8.60e-01 Pobj: -9.3197236e-01 Ad: 1.00e+00 Dobj: -9.3290379e-01
Stuck at edge of primal feasibility, giving up.
Partial Success: SDP solved with reduced accuracy
Primal objective value: -9.3197289e-01
Dual objective value: -9.3290351e-01
Relative primal infeasibility: 4.73e-08
Relative dual infeasibility: 6.44e-10
Real Relative Gap: -3.25e-04
XZ Relative Gap: 1.04e-08
DIMACS error measures: 6.21e-08 0.00e+00 1.70e-09 0.00e+00 -3.25e-04 1.04e-08
value(t) = -0.9319728948556198This stuck at edge of primal feasibility seems strange. SCS solves the problem in my implementation without a hitch: https://github.com/kalmarek/SymbolicWedderburn.jl/runs/4854467386?check_suite_focus=true#step:6:241 why is CSDP struggling here?
blegat
This is ready, but since SymbolicWedderburn-0.3 is not released it will fail ;)