Re: [igraph] Can't get same results from newer layout.fruchterman.reingold

[HIGUCHI Koichi]

Oh, now I understand.
I really appreciate your help.

Igraph is great software. I am enjoying using it and it helps me and
lot of scholars to figure out complex data structures.

Thank you!

2014-02-12 1:01 GMT+09:00 Gábor Csárdi <address@hidden>:
> On Tue, Feb 11, 2014 at 10:56 AM, HIGUCHI Koichi <address@hidden>
> wrote:
>>
>> Hi, thank you very much for your reply!
>>
>> I use "start" option in "layout.fruchterman.reingold()" command in the
>> test code to avoid random start. So "set.seed()" does not have any
>> effect on the results, does it?
>
>
> The algorithm is randomized, in the sense that tiny random displacement
> values might be added to the coordinates in each iteration. So you need to
> set the seed to get the same result, although if your graph is not too
> dense, then you might get away without randomization.
>
>> I guess the algorism has been improved and I need to use old version
>> to get the same result. Maybe I can run secondary R instance with
>> older Igraph to do layout, or just move to the improved layout
>> function.
>
>
> You can install another version of the same package into a different R
> library directory, you just cannot load them at the same time.
>
> Gabor
>
>>
>> Thanks again.
>> Best regards.
>>
>> 2014-02-12 0:39 GMT+09:00 Gábor Csárdi <address@hidden>:
>> > Hi, you also need to specify the random seed via set.seed.
>> >
>> > That said, you still might get different layouts with different igraph
>> > versions, because we are trying to improve the algorithms. If setting
>> > the
>> > seed does not work, then you can generate the layout in the old version,
>> > save the coordinates and then use them with the new version.
>> >
>> > Unfortunately R does not really support using different versions of the
>> > same
>> > package at the same time.
>> >
>> > Best,
>> > Gabor
>> >
>> >
>> > On Tue, Feb 11, 2014 at 10:35 AM, HIGUCHI Koichi
>> > <address@hidden>
>> > wrote:
>> >>
>> >> Does anybody have any ideas?
>> >>
>> >> Thanks.
>> >>
>> >> 2013-08-29 15:32 GMT+09:00 HIGUCHI Koichi <address@hidden>:
>> >> > Hello.
>> >> >
>> >> > I am using igraph R package and moving from igraph 0.5.5 to 0.6.5.
>> >> >
>> >> > But I found that I can't get the same result from
>> >> > layout.fruchterman.reingold function of 0.6.5 even if I specified the
>> >> > same network object and the same starting positions. The output
>> >> > layouts are different between 0.5.5 and 0.6.5.
>> >> >
>> >> > Is there any way to reproduce the same result as 0.5.5 using 0.6.5??
>> >> >
>> >> > Here is my test code:
>> >> > http://koichi.nihon.to/psnl/tmp/net_test.r
>> >> >
>> >> > igraph 0.5.5 result:
>> >> > http://koichi.nihon.to/psnl/tmp/055.png
>> >> >
>> >> > igraph 0.6.5 result:
>> >> > http://koichi.nihon.to/psnl/tmp/065.png
>> >> >
>> >> > New layout is not bad at all, but I would like to have the same
>> >> > results when I input the same data.
>> >> >
>> >> > Thank you! And sorry for my poor English.
>> >> >
>> >> > BTW, I am developing a content analysis (text data mining) software
>> >> > using the igraph package. Igraph package is really useful and nice
>> >> > software! Thank you very much for the effort!
>> >> > http://khc.sourceforge.net/en/gallery/
>> >> >
>> >> > Koichi, Higuchi
>> >>
>> >> _______________________________________________
>> >> igraph-help mailing list
>> >> address@hidden
>> >> https://lists.nongnu.org/mailman/listinfo/igraph-help
>> >
>> >
>> >
>> > _______________________________________________
>> > igraph-help mailing list
>> > address@hidden
>> > https://lists.nongnu.org/mailman/listinfo/igraph-help
>>
>> _______________________________________________
>> igraph-help mailing list
>> address@hidden
>> https://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/igraph-help
>



-- 
立命館大学 産業社会学部 准教授
  樋口 耕一
    〒603-8577 京都市北区等持院北町56-1
    Tel: 075-466-3081
    E-Mail: address@hidden
    Web:    http://koichi.nihon.to/psnl