[Cbc] results differ between log and printed result

[James Howey]

Please to ignore this.

 

I was manipulating readlp and failing to load my last SOS.

 

All by design.

 

Sorry for the interruption.

 

jkh

 

From: James Howey [mailto:james.howey at gmail.com] 
Sent: Monday, May 21, 2018 1:18 PM
To: Cbc at list.coin-or.org
Subject: results differ between log and printed result

 

I solve a problem with preprocess on. The reported objective seems sensible.

 

Cgl0013I Postprocessed model is infeasible - possible tolerance issue - try
without preprocessing

2109 bounds tightened after postprocessing

 

 

Result - Stopped on time limit

 

Objective value:                -479424.31287049

Lower bound:                    -788211.087

Gap:                            0.39

Enumerated nodes:               24677

Total iterations:               1884812

Time (CPU seconds):             701.38

Time (Wallclock seconds):       701.38

 

Total time (CPU seconds):       702.91   (Wallclock seconds):       702.91

 

But consult the printed solution file and see a different, and incorrect,
figure for the objective.

 

Stopped on time - objective value -16822582.00128270

      0 E_worth(29)                               16822582
-1

      1 E_draw(0)                                  -100000
0

      2 E_TX_draw(0)                               -100000
0

 

Presumably, I am seeing the postprocessed solution

 

Another example. Zero is the correct optimum:

 

Coin0505I Presolved problem not optimal, resolve after postsolve

Cgl0013I Postprocessed model is infeasible - possible tolerance issue - try
without preprocessing

2105 bounds tightened after postprocessing

 

91 relaxed row infeasibilities - summing to 2.5858e+06

420 relaxed row infeasibilities - summing to 5.11949e+09

308 relaxed row infeasibilities - summing to 2.30758e+09

 

Result - Optimal solution found

 

Objective value:                0.00000000

Enumerated nodes:               322

Total iterations:               25518

Time (CPU seconds):             61.13

Time (Wallclock seconds):       61.13

 

Total time (CPU seconds):       62.66   (Wallclock seconds):       62.66

 

Solution file looks like this.

 

Optimal - objective value 1008468.84223082

      6 E_shrt(6)                               -53235.737
-1.110223e-16

      7 E_shrt(7)                                   -57923
0

     10 E_shrt(10)                                  -57923
0

 

I can provide the model on request.

 

This is latest trunk build (with or without the new special flag set)

 

I can't be certain when this began, as I only recently tried preprocess
again (which is awesome).

 

Is this possibly by design?

 

Is there a way to turn postprocess off?

 

Thanks,

 

jkh

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20180522/e72eca39/attachment.html>