[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-glpk] Fwd: Re: Infinite cycles
From: |
Heinrich Schuchardt |
Subject: |
Re: [Help-glpk] Fwd: Re: Infinite cycles |
Date: |
Tue, 17 Jan 2017 00:12:18 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Icedove/45.4.0 |
On 01/16/2017 11:59 PM, Andrew Makhorin wrote:
>
>> the interesting thing about Matheu's example is that the infeasability
>> is constantly increasing over multiple orders of magnitude.
>
> I'm unable to reproduce the effect. Glpsol with default options has no
> problem on solving Mathieu's example. If --norelax option is specified,
> the primal simplex falls into infinite loop, but this might be expected,
> because Harris' ratio test (that is, --relax option used by default)
> decreases the number of degenerate steps and thus prevents cycling in
> many cases. If the primal simplex fails, I'd recommend using --dual and
> --flip options.
>
>
> Andrew Makhorin
>
>
With GLPK 4.61 on Linux 64bit and --norelax it is not simply an infinite
loop. The infeasibility increases constantly as shown in the log below.
glpsol -m GLP_buggy_6d21.mod --norelax
* 500: obj = 1.262121212e+04 inf = 4.004e-15 (2)
* 1000: obj = 1.262121212e+04 inf = 1.052e-14 (2)
* 1500: obj = 1.262121212e+04 inf = 1.703e-14 (2)
* 2000: obj = 1.262121212e+04 inf = 2.355e-14 (2)
* 2500: obj = 1.262121212e+04 inf = 3.006e-14 (2)
* 3000: obj = 1.262121212e+04 inf = 3.658e-14 (2)
* 3500: obj = 1.262121212e+04 inf = 4.309e-14 (2)
* 4000: obj = 1.262121212e+04 inf = 4.961e-14 (2)
* 4500: obj = 1.262121212e+04 inf = 5.612e-14 (2)
* 5000: obj = 1.262121212e+04 inf = 6.264e-14 (2)
* 5500: obj = 1.262121212e+04 inf = 6.915e-14 (2)
* 6000: obj = 1.262121212e+04 inf = 7.567e-14 (2)
* 6500: obj = 1.262121212e+04 inf = 8.218e-14 (2)
* 7000: obj = 1.262121212e+04 inf = 8.870e-14 (2)
* 7500: obj = 1.262121212e+04 inf = 9.521e-14 (2)
* 8000: obj = 1.262121212e+04 inf = 1.017e-13 (2)
...
*177000: obj = 1.262121211e+04 inf = 2.304e-12 (2)
*177500: obj = 1.262121211e+04 inf = 2.310e-12 (2)
*178000: obj = 1.262121211e+04 inf = 2.317e-12 (2)
*178500: obj = 1.262121211e+04 inf = 2.323e-12 (2)
*179000: obj = 1.262121211e+04 inf = 2.330e-12 (2)
...
*826000: obj = 1.262121208e+04 inf = 1.076e-11 (2)
*826500: obj = 1.262121208e+04 inf = 1.077e-11 (2)
*827000: obj = 1.262121208e+04 inf = 1.077e-11 (2)
*827500: obj = 1.262121208e+04 inf = 1.078e-11 (2)
*828000: obj = 1.262121208e+04 inf = 1.079e-11 (2)
...
*10719000: obj = 1.262121156e+04 inf = 1.397e-10 (2)
*10719500: obj = 1.262121156e+04 inf = 1.397e-10 (2)
*10720000: obj = 1.262121156e+04 inf = 1.397e-10 (2)
*10720500: obj = 1.262121156e+04 inf = 1.397e-10 (2)
*10721000: obj = 1.262121156e+04 inf = 1.397e-10 (2)
- [Help-glpk] Infinite cycles, Mathieu Dutour, 2017/01/11
- Re: [Help-glpk] Infinite cycles, Heinrich Schuchardt, 2017/01/11
- Re: [Help-glpk] Infinite cycles, Andrew Makhorin, 2017/01/11
- Re: [Help-glpk] Infinite cycles, Mathieu Dutour, 2017/01/12
- Re: [Help-glpk] Infinite cycles, Mathieu Dutour, 2017/01/16
- [Help-glpk] Fwd: Re: Infinite cycles, Heinrich Schuchardt, 2017/01/16
- Re: [Help-glpk] Fwd: Re: Infinite cycles, Andrew Makhorin, 2017/01/16
- Re: [Help-glpk] Fwd: Re: Infinite cycles,
Heinrich Schuchardt <=
- Re: [Help-glpk] Fwd: Re: Infinite cycles, Mathieu Dutour, 2017/01/17
- Re: [Help-glpk] Fwd: Re: Infinite cycles, Andrew Makhorin, 2017/01/17
- Re: [Help-glpk] Fwd: Re: Infinite cycles, Mathieu Dutour, 2017/01/17