[Cbc] Cbc Digest, Vol 96, Issue 1
Alexis
aguigue at softree.com
Thu Jun 11 13:29:05 EDT 2015
Hi John
I managed to make mipStart works with some slight code modifications ad
you suggested. Thanks again.
I am doing some evaluation now.
Cheers
Alexis
On 2015-06-02 3:21 PM, cbc-request at list.coin-or.org wrote:
> Send Cbc mailing list submissions to
> cbc at list.coin-or.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> http://list.coin-or.org/mailman/listinfo/cbc
> or, via email, send a message with subject or body 'help' to
> cbc-request at list.coin-or.org
>
> You can reach the person managing the list at
> cbc-owner at list.coin-or.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Cbc digest..."
>
>
> Today's Topics:
>
> 1. Re: Cbc Digest, Vol 95, Issue 11 (Alexis)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Mon, 01 Jun 2015 14:50:01 -0600
> From: Alexis <aguigue at softree.com>
> To: cbc at list.coin-or.org
> Subject: Re: [Cbc] Cbc Digest, Vol 95, Issue 11
> Message-ID: <556CC579.7050800 at softree.com>
> Content-Type: text/plain; charset="windows-1252"; Format="flowed"
>
> Hi John
>
> Thank you very much for your help.
>
> Before you replied, I figured that it would make way more sense to fix
> the l_xx_xx rather than the u's so I changed my solini.txt file
> accordingly. It only contains these variables now. Once these are set,
> the remaining problem in Sample.lp is entirely an lp
> as the SOS2 constraints are satisfied.
>
> Now about feasibility, this is interesting. Solini.txt has been
> generated using the problem in SampleIni.lp. This problem
> is just a subset in terms of constraints and variables of the original
> problem Sample.lp with a different objective function.
> The added variables/constraints should not create infeasibility (it is a
> flow problem almost entirely independent) so I suspect
> sensitivity to exact values is the issue. If the issue repeats with
> l_xx_xx being fixed, then you are right, I should probably try to bypass
> the file.
>
> So Sample.lp is the same file as before. SampleIni.lp is the file that
> has been used to generate solini.txt.
>
> Alexis
>
> On 2015-05-31 10:00 AM, cbc-request at list.coin-or.org wrote:
>> Send Cbc mailing list submissions to
>> cbc at list.coin-or.org
>>
>> To subscribe or unsubscribe via the World Wide Web, visit
>> http://list.coin-or.org/mailman/listinfo/cbc
>> or, via email, send a message with subject or body 'help' to
>> cbc-request at list.coin-or.org
>>
>> You can reach the person managing the list at
>> cbc-owner at list.coin-or.org
>>
>> When replying, please edit your Subject line so it is more specific
>> than "Re: Contents of Cbc digest..."
>>
>>
>> Today's Topics:
>>
>> 1. Re: Cbc Digest, Vol 95, Issue 7 (John Forrest)
>>
>>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Sun, 31 May 2015 15:36:46 +0100
>> From: John Forrest <john.forrest at fastercoin.com>
>> To: cbc at list.coin-or.org
>> Subject: Re: [Cbc] Cbc Digest, Vol 95, Issue 7
>> Message-ID: <556B1C7E.7050801 at fastercoin.com>
>> Content-Type: text/plain; charset="windows-1252"; Format="flowed"
>>
>> Alexis,
>>
>> It was just good luck that it worked.
>>
>> The partial solution in solini.txt does not seem to be a valid
>> solution. If I fix those variables the problem is infeasible (the same
>> is true if I just fix a's or u's). It looks as if the problem is very
>> sensitive to exacat values and that printed values were not good enough.
>>
>> The code as-is fixes other variables to zero, finds problem is
>> infeasible and then says - OK lets free up all non-integer variables to
>> see if that works i.e. free all variables. It then gets root solution -
>> which in this case is valid.
>>
>> I have modified code (on a compile time option) so that it will fix all
>> variables in solini.txt and then do a mini branch and bound - that
>> should work if solini.txt had a valid solution.
>>
>> Your best bet is to code your own version - fix some to known value and
>> do branch and bound on easier problem stopping at first solution and
>> then passing that in without loss of accuracy. It looks as if fixing
>> l_xx_xx to zero is much more forgiving that fixing u. Have you got
>> something I could test on?
>>
>> John
>> On 27/05/15 19:14, Alexis wrote:
>>> Hi John,
>>>
>>> I managed to implement something and it seems to work (?).
>>>
>>> Here is what I have done is somewhat pseudocode:
>>>
>>> OsiClpSolverInterface mySolver;
>>> Set Variables, Constraints and SOS2 Constraints in mySolver (see
>>> Sample.lp file for the output)
>>> CbcModel myModel(mySolver);
>>> const char * argv2[] = { "cbc", "-directory", mydirectory,
>>> "-mipStart", "solini.txt", "-solve", "-quit" }; where solini.txt is a
>>> valid (partial) initial solution
>>> CbcMain1(7, argv2, myModel);
>>>
>>> I am saying it seems to work because the function
>>> computeCompleteSolution(...) in CbcMipStartIO.cpp is successfully
>>> called (returns 0) and the following code
>>> is called (in CbcSolver.cpp)
>>>
>>> int status = computeCompleteSolution( babModel_,
>>> colNames, mipStart, &x[0], obj );
>>> if (!status) {
>>> babModel_->setBestSolution( &x[0],
>>> static_cast<int>(x.size()), obj, false );
>>> babModel_->setSolutionCount(1);
>>> }
>>>
>>> Now a few words about the problem. Here is a typical SOS2 constraints
>>> (see Sample.lp file)
>>>
>>> SOS2CONS1: S2:: l_1_1:1 l_1_2:2 l_1_3:3 l_1_4:4 l_1_5:5 l_1_6:6
>>> l_1_7:7 l_1_8:8 l_1_9:9 l_1_10:10 l_1_11:11 l_1_12:12 l_1_13:13
>>> l_1_14:14 l_1_15:15 l_1_16:16 l_1_17:17 l_1_18:18 l_1_19:19 l_1_20:20
>>> l_1_21:21 l_1_22:22 l_1_23:23 l_1_24:24 l_1_25:25 l_1_26:26 l_1_27:27
>>> l_1_28:28 l_1_29:29 l_1_30:30 l_1_31:31 l_1_32:32 l_1_33:33 l_1_34:34
>>>
>>> together with
>>>
>>> COVER_1: l_1_1 + l_1_2 + l_1_3 + l_1_4 + l_1_5 + l_1_6 + l_1_7 +
>>> l_1_8 + l_1_9 + l_1_10
>>> + l_1_11 + l_1_12 + l_1_13 + l_1_14 + l_1_15 + l_1_16 + l_1_17 +
>>> l_1_18 + l_1_19 + l_1_20
>>> + l_1_21 + l_1_22 + l_1_23 + l_1_24 + l_1_25 + l_1_26 + l_1_27 +
>>> l_1_28 + l_1_29 + l_1_30
>>> + l_1_31 + l_1_32 + l_1_33 + l_1_34 = 1
>>>
>>> and
>>>
>>> INTER_1: - u_1 -10.44566 l_1_1 -9.82066 l_1_2 -9.19566 l_1_3 -8.57066
>>> l_1_4 -7.94566 l_1_5 -7.32066 l_1_6 -6.69566 l_1_7 -6.07066 l_1_8
>>> -5.44566 l_1_9
>>> -4.82066 l_1_10 -4.19566 l_1_11 -3.57066 l_1_12 -2.94566 l_1_13
>>> -2.32066 l_1_14 -2.25232 l_1_15 -1.69566 l_1_16 -1.07066 l_1_17
>>> -0.61982 l_1_18 -0.44566 l_1_19
>>> + 0.17934 l_1_20 + 0.26541 l_1_21 + 0.80434 l_1_22 + 1.42934 l_1_23 +
>>> 2.05434 l_1_24 + 2.67934 l_1_25 + 3.30434 l_1_26 + 3.92934 l_1_27 +
>>> 4.55434 l_1_28 + 5.17934 l_1_29
>>> + 5.80434 l_1_30 + 6.42934 l_1_31 + 7.05434 l_1_32 + 7.67934 l_1_33 +
>>> 8.30434 l_1_34 = 0
>>>
>>> In "solini.txt", the value for u_1 is given so that sets the l_1_i
>>> variables.
>>>
>>> Am I on the right track? I am still not sure 100% about what mipStart
>>> does. Is it just calculating and setting up a feasible solution (and
>>> an upper bound) for the B&B?
>>>
>>> Thanks again, I will keep you posted about my investigations.
>>>
>>> Alexis
>>>
>>> On 2015-05-20 7:27 AM, cbc-request at list.coin-or.org wrote:
>>>> Send Cbc mailing list submissions to
>>>> cbc at list.coin-or.org
>>>>
>>>> To subscribe or unsubscribe via the World Wide Web, visit
>>>> http://list.coin-or.org/mailman/listinfo/cbc
>>>> or, via email, send a message with subject or body 'help' to
>>>> cbc-request at list.coin-or.org
>>>>
>>>> You can reach the person managing the list at
>>>> cbc-owner at list.coin-or.org
>>>>
>>>> When replying, please edit your Subject line so it is more specific
>>>> than "Re: Contents of Cbc digest..."
>>>>
>>>>
>>>> Today's Topics:
>>>>
>>>> 1. Re: Cbc Digest, Vol 95, Issue 6 (Alexis)
>>>> 2. Re: Cbc Digest, Vol 95, Issue 6 (John Forrest)
>>>> 3. Re: Cbc Digest, Vol 95, Issue 6 (Haroldo Gambini Santos)
>>>>
>>>>
>>>> ----------------------------------------------------------------------
>>>>
>>>> Message: 1
>>>> Date: Tue, 19 May 2015 12:56:16 -0600
>>>> From: Alexis<aguigue at softree.com>
>>>> To:cbc at list.coin-or.org
>>>> Subject: Re: [Cbc] Cbc Digest, Vol 95, Issue 6
>>>> Message-ID:<555B8750.3010300 at softree.com>
>>>> Content-Type: text/plain; charset="windows-1252"; Format="flowed"
>>>>
>>>> Thanks John
>>>>
>>>> I have done preliminary testing of mipStart using standalone cbc. It
>>>> looks promising.
>>>> However, our real problems have a bunch of SOS2 constraints. I checked
>>>> but it does not
>>>> seems that cbc v2.9.0 can import/read such files yet, am I correct? The
>>>> other
>>>> thing I was wondering is that whether mipStart would work with SOS2
>>>> constraints?
>>>> Say we have a lp problem with binary variables and some SOS2
>>>> constraints, would
>>>> mipStart work by, again, just providing the values of the binary
>>>> variables, and
>>>> leaving the variables in the SOS2 constraints to be decided?
>>>>
>>>> Also, is there some doc/pub about Haroldo implementation/method?
>>>>
>>>> Thanks in advance
>>>>
>>>> Alexis
>>>>
>>>> On 2015-05-13 10:00 AM,cbc-request at list.coin-or.org wrote:
>>>>> Send Cbc mailing list submissions to
>>>>> cbc at list.coin-or.org
>>>>>
>>>>> To subscribe or unsubscribe via the World Wide Web, visit
>>>>> http://list.coin-or.org/mailman/listinfo/cbc
>>>>> or, via email, send a message with subject or body 'help' to
>>>>> cbc-request at list.coin-or.org
>>>>>
>>>>> You can reach the person managing the list at
>>>>> cbc-owner at list.coin-or.org
>>>>>
>>>>> When replying, please edit your Subject line so it is more specific
>>>>> than "Re: Contents of Cbc digest..."
>>>>>
>>>>>
>>>>> Today's Topics:
>>>>>
>>>>> 1. Re: Cbc Digest, Vol 95, Issue 5 (Alexis)
>>>>> 2. Re: Cbc Digest, Vol 95, Issue 5 (John Forrest)
>>>>>
>>>>>
>>>>> ----------------------------------------------------------------------
>>>>>
>>>>> Message: 1
>>>>> Date: Tue, 12 May 2015 11:09:58 -0600
>>>>> From: Alexis<aguigue at softree.com>
>>>>> To:cbc at list.coin-or.org
>>>>> Subject: Re: [Cbc] Cbc Digest, Vol 95, Issue 5
>>>>> Message-ID:<555233E6.8030501 at softree.com>
>>>>> Content-Type: text/plain; charset="windows-1252"; Format="flowed"
>>>>>
>>>>> Thanks John.
>>>>>
>>>>> I am actually using the API. I found this example
>>>>>
>>>>> https://projects.coin-or.org/Cbc/browser/trunk/Cbc/examples/hotstart.cpp
>>>>>
>>>>> I will try it. Is the example you were referring to?
>>>>>
>>>>> On 2015-05-12 10:00 AM,cbc-request at list.coin-or.org wrote:
>>>>>> Send Cbc mailing list submissions to
>>>>>> cbc at list.coin-or.org
>>>>>>
>>>>>> To subscribe or unsubscribe via the World Wide Web, visit
>>>>>> http://list.coin-or.org/mailman/listinfo/cbc
>>>>>> or, via email, send a message with subject or body 'help' to
>>>>>> cbc-request at list.coin-or.org
>>>>>>
>>>>>> You can reach the person managing the list at
>>>>>> cbc-owner at list.coin-or.org
>>>>>>
>>>>>> When replying, please edit your Subject line so it is more specific
>>>>>> than "Re: Contents of Cbc digest..."
>>>>>>
>>>>>>
>>>>>> Today's Topics:
>>>>>>
>>>>>> 1. Hot Start for MIP (Alexis)
>>>>>> 2. Re: Hot Start for MIP (John Forrest)
>>>>>>
>>>>>>
>>>>>> ----------------------------------------------------------------------
>>>>>>
>>>>>> Message: 1
>>>>>> Date: Mon, 11 May 2015 11:07:35 -0600
>>>>>> From: Alexis<aguigue at softree.com>
>>>>>> To:cbc at list.coin-or.org
>>>>>> Subject: [Cbc] Hot Start for MIP
>>>>>> Message-ID:<5550E1D7.8030902 at softree.com>
>>>>>> Content-Type: text/plain; charset="utf-8"; Format="flowed"
>>>>>>
>>>>>> I have got a large MIP to solve for which I would like to implement a
>>>>>> hot start strategy.
>>>>>>
>>>>>> My understanding is that I should use the function
>>>>>> CbcModel::setHotstartSolution(const double *solution, const int
>>>>>> *priorities=NULL),
>>>>>> where solution assumes the same ''order'' of variables as defined for
>>>>>> the model itself. I am not so sure to understand the meaning of
>>>>>> the pointer priorities though. The hotstart solution I intend to give is
>>>>>> an actual feasible solution to the MIP.
>>>>>>
>>>>>> Am I on the right track?
>>>>>>
>>>>>> Thanks in advance for the help
>>>>>>
>>> --
>>> Alexis Guigue, E.I.T., B.Eng., M.A.Sc., Ph.D.
>>> Software Engineer/Research Engineer
>>> aguigue at softree.com
>>> 604-519-6222, ext. 3
>>>
>>> Softree Technical Systems Inc. <http://www.softree.com>
>>> Engineering an Easier Way
>>>
>>> Softree News & Updates <http://http://eepurl.com/XJT-z>
>>>
>>>
>>> _______________________________________________
>>> Cbc mailing list
>>> Cbc at list.coin-or.org
>>> http://list.coin-or.org/mailman/listinfo/cbc
>> -------------- next part --------------
>> An HTML attachment was scrubbed...
>> URL: <http://list.coin-or.org/pipermail/cbc/attachments/20150531/890bd549/attachment-0001.html>
>>
>> ------------------------------
>>
>> _______________________________________________
>> Cbc mailing list
>> Cbc at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/cbc
>>
>>
>> End of Cbc Digest, Vol 95, Issue 11
>> ***********************************
>>
>
--
Alexis Guigue, E.I.T., B.Eng., M.A.Sc., Ph.D.
Software Engineer/Research Engineer
aguigue at softree.com
604-519-6222, ext. 3
Softree Technical Systems Inc. <http://www.softree.com>
Engineering an Easier Way
Softree News & Updates <http://http://eepurl.com/XJT-z>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20150611/caff76bb/attachment-0001.html>
More information about the Cbc
mailing list