[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-glpk] glpk 4.9 release information
From: |
Andrew Makhorin |
Subject: |
Re: [Help-glpk] glpk 4.9 release information |
Date: |
Thu, 26 Jan 2006 13:49:04 +0300 |
>> The lower bound obtained from a relaxation is only a bound. Its value
>> may depend, for example, on transformations performed by presolver or
>> on cutting planes added to the relaxation. So it does not express any
>> essential information.
> I agree with you, but when using cplex I use it as an indication of the
> progress of the solver. If the lower bound only changes in, say, the 12th
> digit it is quite likely that the solver will run out of memory before the
> optimal solution is found. So if it would be possible to optionally print
> it I would most likely use it for this kind of thing.
In most cases the lower bound does not increases smoothly; it keeps
its value until corresponding node has been pruned, so it is not a good
indicator of the progress. In glpk a better indication is the number of
active and fathomed nodes displayed in parentheses. If the number of
active nodes increases much faster than the number of fathomed nodes for
a long time (say, one hour), the problem is hard for glpk solver.