Re: Tolerance in LinearProgramming
- To: mathgroup at smc.vnet.net
- Subject: [mg98844] Re: Tolerance in LinearProgramming
- From: dh <dh at metrohm.com>
- Date: Mon, 20 Apr 2009 05:41:15 -0400 (EDT)
- References: <gsc026$5nj$1@smc.vnet.net>
Hi Slava,
I do not know how mathematica does the calculation. But because a>=0 is
equivalent to !a<0, mathematica may use the second possibility, what
would explain the behaviour of tolerance.
Anyway, it is always better to calculate with exact numbers (Integers or
Rationals), then you do not have these problems.
Daniel
Slava Rychkov wrote:
> Hi!
>
> I would like to understand better the meaning of the Tolerance parameter
> in LinearProgramming. For example, I would like to understand why
>
> M = {{1000.,
> 300., -300.0}, {10., -2000., -1.}, {100000., -1000., -90.}};
> LinearProgramming[{0, 0, 0}, M, {0, 0, 0}, {1, 1, 1},
> Tolerance -> 0.1]
>
> does not find a solution, while with smaller Tolerance it does:
>
> LinearProgramming[{0, 0, 0}, M, {0, 0, 0}, {1, 1, 1},
> Tolerance -> 0.01]
>
> Naively, I find this behavior counterintuitive.
> Thanks a lot,
>