[Dip-tickets] [Dip] #59: (MILPBlock) "Point violates row" with large relative violation

Dip coin-trac at coin-or.org
Fri Dec 3 01:45:18 EST 2010


#59: (MILPBlock) "Point violates row" with large relative violation
-------------------+--------------------------------------------------------
Reporter:  vejlin  |       Type:  defect
  Status:  new     |   Priority:  major 
 Version:  trunk   |   Keywords:        
-------------------+--------------------------------------------------------
 I am getting some rather large violations with the unmodified MILPBlock
 example:
 "Point violates row 2 -> r(2) LB= 0.0000000 ax= -1017.0750594 UB=  INF
 RelViol= 1017.0750594"

 This happened with an unmodified MILPBlock example, the "harp2" instance
 from miplib and the block file (list format):
 "0 74
 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 106 111"

 Constraints 0-71 and 111 are column disjount Special Ordeded Sets (\sum
 x_i  = 1) and constraint 106 is a degenerated 0-1 Knapsack constraint
 (capacity = 0) covering only a subset of the columns covered by the SOS
 constraints. I am auto generating my block files so it may not be a very
 good decomposition, but MILPBlock responds oddly and I cannot figure out
 why.
 The knapsack constraint simply fixes some variables to zero, and the
 remaining SOS constraints should not be that problematic. The problem is
 bounded, if nothing else then because all variables are binary.

-- 
Ticket URL: <https://projects.coin-or.org/Dip/ticket/59>
Dip <https://projects.coin-or.org/Dip>
An extensible software framework for implementing decompositon-based bounding algorithms for use in solving large-scale discrete optimization problems.



More information about the Dip-tickets mailing list