Re: [Help-glpk] glpk 4.59.2 non-official test release

[Chris Matrakidis]

Andrew,

> I tried solving it in the following way. First I used --bestp to find a
> feasible solution quickly and save it to a file:
...
> And then I used that solution as an incumbent value trying to "prove"
> its optimality:

This is essentially what --proxy does when searching for a first
solution. However this doesn't work now, while in the previous version
it was finding solutions. I added the line:
#define glp_term_out(x) x
at the beginning of proxy.c and found this:

[...]
Applying PROXY heuristic...
Proxy's time limit set to 60 seconds.
Proxy's relative improvement set to 1.00 %.
Searching for a feasible solution...
GLPK Simplex Optimizer, v4.59
394 rows, 321 columns, 1213 non-zeros
*     0: obj =  8.685404001e+004 inf =  0.000e+000 (0)
OPTIMAL LP SOLUTION FOUND
GLPK Integer Optimizer, v4.59
394 rows, 321 columns, 1213 non-zeros
154 integer variables, all of which are binary
Integer optimization begins...
+     0: mip =     not found yet >=              -inf        (1; 0)
Warning: numerical instability (dual simplex, phase II)
 242500: obj =  2.414880323e+009 inf =  7.186e+007 (25)
 243000: obj =  2.414880323e+009 inf =  7.186e+007 (25)
 243500: obj =  2.414880323e+009 inf =  7.186e+007 (25)
 244000: obj =  2.414880323e+009 inf =  7.186e+007 (25)
[...]

This was with just "glpsol --proxy water.mps" and is essentially the
same failure as the one I reported earlier:

>> + 86090: mip =     not found yet >=  1.160667584e+005        (3265; 251)
>> Warning: numerical instability (dual simplex, phase II)
>>  294000: obj =  3.121033256e+008 inf =  5.425e+006 (14)
>>  294500: obj =  3.121033256e+008 inf =  5.425e+006 (14)
>>  295000: obj =  3.121033256e+008 inf =  5.425e+006 (14)
>>  295500: obj =  3.121033256e+008 inf =  5.425e+006 (14)
>>  296000: obj =  3.121033256e+008 inf =  5.425e+006 (14)
>>  296500: obj =  3.121033256e+008 inf =  5.425e+006 (14)


Best Regards,

Chris Matrakidis