[Scip] Using SCIP included in another source code to solve quadratic constraints
Jiehua Chen
jiehua.chen at tu-berlin.de
Mon Apr 20 11:16:52 CEST 2015
Hi,
I still can't grasp the crucial part of the problem. I made the
following by compiling:
tar xvf scipoptsuite-3.1.1.tgz
cd scipoptsuite-3.1.1/
cd scip-3.1.1/
make ZIMPL=false IPOPT=true clean
make ZIMPL=false IPOPT=true links
For .so, I didn't give any path.
For ipopt I gave /usr
make ZIMPL=false IPOPT=true
Did I miss something?
Best,
Jiehua
On 04/14/2015 04:02 PM, Stefan Vigerske wrote:
> Hi,
>
> run SCIP with some timelimit and look at the final solve statistics. Did
> the "subnlp" heuristic run? If not, add a #define SCIP_DEBUG on the
> (very) top of heur_subnlp.c to get some debugging output from this
> heuristic. It may tell you why it didn't run or didn't find a feasible
> point.
>
> Stefan
>
> On 04/14/2015 11:21 AM, Jiehua Chen wrote:
>> I did add mumps and some other stuff in make. However, it still runs
>> very slowly. What kind of log messages do you need?
>>
>> On 04/14/2015 10:25 AM, Stefan Vigerske wrote:
>>> Hi,
>>>
>>> something that people miss sometimes is that Ipopt needs a linear solver
>>> to work, e.g., Mumps. It doesn't complain when there is noone at compile
>>> time, as it can load it dynamically during runtime.
>>> But if make test of Ipopt worked fine, that shouldn't be the issue.
>>> From the few log lines you are printing there, I don't see what could
>>> be the issue.
>>>
>>> Stefan
>>>
>>> On 04/14/2015 10:19 AM, Jiehua Chen wrote:
>>>> Hi,
>>>>
>>>> Yes. The problem was that the symbolic link should start from some
>>>> other
>>>> directory than the one when you "make". The hint in the INSTALL was not
>>>> very helpful. Nevertheless, I still don't get the result after 10 hours
>>>> (using ipopt):
>>>>
>>>>
>>>>
>>>> ******************************************************************************
>>>>
>>>>
>>>>
>>>>
>>>> This program contains Ipopt, a library for large-scale nonlinear
>>>> optimization.
>>>> Ipopt is released as open source code under the Eclipse Public
>>>> License
>>>> (EPL).
>>>> For more information visit http://projects.coin-or.org/Ipopt
>>>> ******************************************************************************
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> 0.0s| 1 | 2 | 0 | - | 220k| 0 | 0 | 12 | 9 |
>>>> 12 | 0 | 0 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 0.4s| 100 | 91 | 662 | 6.7 | 307k| 55 | - | 12 | 9 |
>>>> 0 | 0 | 633 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 0.5s| 200 | 171 | 1503 | 7.6 | 403k| 81 | 0 | 12 | 9 |
>>>> 12 | 0 |1222 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 0.5s| 300 | 257 | 3265 | 10.9 | 574k| 81 | - | 12 | 9 |
>>>> 12 | 113 |2711 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 0.7s| 400 | 341 | 6199 | 15.5 | 902k| 101 | 0 | 12 | 9 |
>>>> 12 | 93 |5072 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 0.9s| 500 | 437 | 10677 | 21.4 |1138k| 114 | 0 | 12 | 9 |
>>>> 12 | 0 |8380 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>> 1.0s| 600 | 531 | 11329 | 18.9 |1239k| 114 | 0 | 12 | 9 |
>>>> 12 | 98 |8906 | 0 | 0 | 1.000000e+00 | -- | Inf
>>>>
>>>>
>>>> What's going wrong?
>>>>
>>>> Best,
>>>> Jiehua Chen
>>>>
>>>> On 04/13/2015 02:10 PM, Stefan Vigerske wrote:
>>>>> Hi,
>>>>>
>>>>> probably the symlink to the SoPlex source was not set correctly when
>>>>> you
>>>>> called make the first time.
>>>>> Just make sure that lib/spxinc points to the soplex src directory.
>>>>>
>>>>> Stefan
>>>>>
>>>>>
>>>>> On 04/13/2015 01:25 PM, Jiehua Chen wrote:
>>>>>> Hi,
>>>>>>
>>>>>> I encounter the following error message when trying to compile:
>>>>>>
>>>>>> compiling obj/O.linux.x86_64.gnu.opt/lib/lpi/lpi_spx.o
>>>>>> src/lpi/lpi_spx.cpp:76:23: fatal error: spxsolver.h: No such file or
>>>>>> directory
>>>>>> #include "spxsolver.h"
>>>>>>
>>>>>> Can you give me more hints?
>>>>>>
>>>>>> Best
>>>>>> Jiehua
>>>>>>
>>>>>>
>>>>>> On 04/08/2015 03:09 PM, Stefan Vigerske wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> it also takes long for me if I disable the subnlp primal heuristic.
>>>>>>> This
>>>>>>> primal heuristic is also off if you build SCIP without Ipopt. If
>>>>>>> that is
>>>>>>> the case for you, then you might want to consider rebuiling SCIP
>>>>>>> with
>>>>>>> Ipopt (see documentation on hints how to do this).
>>>>>>>
>>>>>>> Best,
>>>>>>> Stefan
>>>>>>>
>>>>>>> On 04/08/2015 02:49 PM, Jiehua Chen wrote:
>>>>>>>> Dear all,
>>>>>>>>
>>>>>>>> I have a quadratic constraint problem which I try to solve by using
>>>>>>>> SCIP
>>>>>>>> (as a library) in C++. However, my program does not terminate
>>>>>>>> after 1
>>>>>>>> hour. Curiously, if I use your online solver
>>>>>>>>
>>>>>>>> http://www.neos-server.org/neos/solvers/go:scip/CPLEX.html
>>>>>>>>
>>>>>>>> , then after several seconds I got the solution. Can you please
>>>>>>>> help
>>>>>>>> me?
>>>>>>>> Thanks!
>>>>>>>>
>>>>>>>> Below is the problem:
>>>>>>>>
>>>>>>>> -------------------------------------------------------------------
>>>>>>>> minimize y#1
>>>>>>>>
>>>>>>>>
>>>>>>>> subject to
>>>>>>>> a#1 = 0
>>>>>>>> a#2 = 0
>>>>>>>>
>>>>>>>> [a#1^2 + a#2^2 - 2 a#1*v#11 - 2 a#2*v#12 - b#1^2 - b#2^2 + 2
>>>>>>>> b#1*v#11 +
>>>>>>>> 2 b#2*v#12] + y#1 <= 0
>>>>>>>> [b#1^2 + b#2^2 - 2 b#1*v#11 - 2 b#2*v#12 - c#1^2 - c#2^2 + 2
>>>>>>>> c#1*v#11 +
>>>>>>>> 2 c#2*v#12] + y#1 <= 0
>>>>>>>> [c#1^2 + c#2^2 - 2 c#1*v#11 - 2 c#2*v#12 - d#1^2 - d#2^2 + 2
>>>>>>>> d#1*v#11 +
>>>>>>>> 2 d#2*v#12] + y#1 <= 0
>>>>>>>>
>>>>>>>> [b#1^2 + b#2^2 - 2 b#1*v#21 - 2 b#2*v#22 - c#1^2 - c#2^2 + 2
>>>>>>>> c#1*v#21 +
>>>>>>>> 2 c#2*v#22] + y#1 <= 0
>>>>>>>> [c#1^2 + c#2^2 - 2 c#1*v#21 - 2 c#2*v#22 - d#1^2 - d#2^2 + 2
>>>>>>>> d#1*v#21 +
>>>>>>>> 2 d#2*v#22] + y#1 <= 0
>>>>>>>> [d#1^2 + d#2^2 - 2 d#1*v#21 - 2 d#2*v#22 - a#1^2 - a#2^2 + 2
>>>>>>>> a#1*v#21 +
>>>>>>>> 2 a#2*v#22] + y#1 <= 0
>>>>>>>>
>>>>>>>> [c#1^2 + c#2^2 - 2 c#1*v#31 - 2 c#2*v#32 - d#1^2 - d#2^2 + 2
>>>>>>>> d#1*v#31 +
>>>>>>>> 2 d#2*v#32] + y#1 <= 0
>>>>>>>> [d#1^2 + d#2^2 - 2 d#1*v#31 - 2 d#2*v#32 - a#1^2 - a#2^2 + 2
>>>>>>>> a#1*v#31 +
>>>>>>>> 2 a#2*v#32] + y#1 <= 0
>>>>>>>> [a#1^2 + a#2^2 - 2 a#1*v#31 - 2 a#2*v#32 - b#1^2 - b#2^2 + 2
>>>>>>>> b#1*v#31 +
>>>>>>>> 2 b#2*v#32] + y#1 <= 0
>>>>>>>>
>>>>>>>> y#1 >= 0.01
>>>>>>>>
>>>>>>>> end
>>>>>>>> -------------------------------------------------------------------
>>>>>>>>
>>>>>>>>
>>>>>>>> Best regards!
>>>>>>>> Jiehua Chen
>>>>>>>> _______________________________________________
>>>>>>>> Scip mailing list
>>>>>>>> Scip at zib.de
>>>>>>>> http://listserv.zib.de/mailman/listinfo/scip
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>
>
>
More information about the Scip
mailing list