jjhforrest
commented
3 years ago Use cbc not clp.
There may be bugs with semi-continuous variables - one output line is
2 semi-continuous variables - report odd behavior
However cbc reports the problem to feasible and gets correct answer. If cbc gives wrong answers to nay of your problems - I will look into it.
clp is getting confused as it does not really know about semi-continuous variables (or any discontinuities).
John Forrest
On 13/11/2020 09:21, raller09 wrote:
Hi,
i am trying to use clp to solve a minimization problem with semi-continuous variables.
Please see my sample problem (i hope I haven't oversimplified it):
|NAME ROWS N R0 L y5050_max L y0954_max E y_T E scv_y3380 E scv_y5050 COLUMNS x3380_p R0 2.8500000000 y0954_max 100.00000000 x3380_p y_T 1.0000000000 scv_y3380 1.0000000000 x5050_p R0 1.7650000000 y5050_max 1.0000000000 x5050_p y_T 1.0000000000 scv_y5050 1.0000000000 x7010_p R0 0.3400000000 y_T 1.0000000000 x3380_s R0 2.8499900000 y0954_max 100.00000000 x3380_s y_T 1.0000000000 scv_y3380 1.0000000000 x5050_s R0 1.7649900000 y5050_max 1.0000000000 x5050_s y_T 1.0000000000 scv_y5050 1.0000000000 x7010_s R0 0.3399900000 y_T 1.0000000000 scv_x3380 scv_y3380 -1.000000000 scv_x5050 scv_y5050 -1.000000000 RHS RHS y5050_max 10.000000000 y0954_max 1.0000000000 RHS y_T 100.00000000 BOUNDS SC BND scv_x3380 LO BND scv_x3380 2.3000000000 SC BND scv_x5050 LO BND scv_x5050 2.3000000000 ENDATA |
I reduced it to 3 (6) variables. Always 2 are in relation to each other (in my example only with different costs) -> _p + _s.
|if a or b are used, then a + b must be 0 or >= value. |
so i added a column scv_y with
|a + b = c -> a + b - c = 0 with c semi-continuous and >= 2.3 |
clp reports this as infeasible:
|D:\Temp\lpsolver\Clp-master-win64-msvc15-md\bin>clp.exe ccc.mps solve Coin LP version devel, build Jul 3 2020 command line - clp.exe ccc.mps solve At line 1 NAME At line 2 ROWS At line 9 COLUMNS At line 22 RHS At line 25 BOUNDS At line 30 ENDATA Problem no_name has 5 rows, 8 columns and 16 elements 2 semi-continuous variables - report odd behavior Model was imported from .\ccc.mps in 0.003 seconds Presolve determined that the problem was infeasible with tolerance of 1e-08 Analysis indicates model infeasible or unbounded Primal infeasible - objective value 43.0495 PrimalInfeasible objective 43.0495 - 3 iterations time 0.002 |
it is even with presolve tolerance set to 0.1 infeasible:
|clp.exe ccc.mps preT 1e-1 solve preTolerance was changed from 1e-08 to 0.1 Presolve determined that the problem was infeasible with tolerance of 0.1 |
if i remove the semi-continuous variables it is feasible. In my example there a only "lower-then" and a var without a constraints. So it should always be solvable.
I've tried a little offset for the equation of "scv_y" ±0.00001 But this leads to an incorrect result in one case.
I tried to solve > 400 sample problems. But t think only 2-3 are problematic.
I am using binarys from bintray for Windows 64 bit (tryed msvc15/16 versions). Should I use the newsgroup or is this the right place?
Am I doing something wrong?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/coin-or/Clp/issues/165, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABWJYHBIK7XBEQCQXVH6MVDSPT3ADANCNFSM4TUK5W2Q.
raller09
Hi,
i am trying to use clp to solve a minimization problem with semi-continuous variables.
Please see my sample problem (i hope I haven't oversimplified it):
I reduced it to 3 (6) variables. Always 2 are in relation to each other (in my example only with different costs) -> _p + _s.
so i added a column scv_y with
clp reports this as infeasible:
it is even with presolve tolerance set to 0.1 infeasible:
if i remove the semi-continuous variables it is feasible. In my example there a only "lower-then" and a var without a constraints. So it should always be solvable.
I've tried a little offset for the equation of "scv_y" ±0.00001 But this leads to an incorrect result in one case.
I tried to solve > 400 sample problems. But t think only 2-3 are problematic.
I am using binarys from bintray for Windows 64 bit (tryed msvc15/16 versions). Should I use the newsgroup or is this the right place?
Am I doing something wrong?