Commit 0ce66249 authored by Gregor Hendel's avatar Gregor Hendel

Add MIP starts and README of Irv Lustig submission

parent 5fbd4b92
The 3 problems enclosed represent a set of sequential solves using
hierarchical optimization, via 3 objectives.
The problem seqsolve.1.mps is the first problem to solve. The objective
is called "shortage". The best known solution has an objective of 4288.
The problem seqsolve.2.short4288.mps is the second problem to solve.
The objective is called "excess". It has a constraint called shortageBound
that adds 0.25 to the objective of the first problem, so that the
new objective is minimized without making the first objective worse.
In the RHS section of the MPS file is the line:
RHS1 shortageBound 4.28825e+03
If a better objective value than 4288 is found for seqsolve.1.mps,
then the above line should be changed.
The best known solution for the second problem (given the objective
bound derived from the first problem) is 384.
The problem seqsolve.3.short4288.excess384.mps is the third problem
to solve. The objective is called "preference". It has 2 constraints,
shortageBound and excessBound that each add 0.25 to the objectives
of the first and second problem, respectively, so that the new
objective is minimized without making either of the first two
objectives worse. In the RHS section of the MPS file are the lines:
RHS1 shortageBound 4.28825e+03
RHS1 excessBound 3.8425e+02
If a better objective value is obtained for the first or second
problems, then these values should be modified.
Note that the last feasible solution for the first problem should be
used as a MIP start for the second problem. Similarly, the last
known feasible solution for the second problem should be used as a
MIP start for the third problem. For convenience, the Gurobi MST
files first.mst and second.mst are supplied, representing the MIP
start to use for seqsolve.2.short4288.mps and
seqsolve.3.short4288.excess384.mps, respectively.
Each of the problems were solved with Gurobi 7.0.0 on a Intel Core
i7-4800MQ CPU @ 2.70GHz, with 4 physical cores, and a total of 8
logical processors. The solver was stopped after 10 hours of
computation for each run. Here is a summary of the results:
Problem Objective Bound Gap
seqsolve.1 4.288000000000e+03 4.275000000000e+03 0.3032%
seqsolve.2.short4288 3.840000000000e+02 1.771829204207e+01 95.3859%
seqsolve.3.short4288.excess384 -3.061900000000e+04 4.245900000000e+04 38.6688%
The gurobi.log file for the 3 solves is also included.
Submitted by Irv Lustig, Princeton Consultants, 11/22/16
This diff is collapsed.
This source diff could not be displayed because it is too large. You can view the blob instead.
This diff is collapsed.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment