Parallelization of LMZ.F3GWR.test in spgwr package

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Parallelization of LMZ.F3GWR.test in spgwr package

Diana Gutiérrez
Dear all,

I was wondering if it would be possible to run the F3 variability test by
Leung et al. (2000) (as included in the spgwr package) using parallel, so
processing times could be reduced when you have a large number of
observations and parameters. I would appreciate any insights and/or code
examples if you happen to know of any that can be shared.

Thanks in advance,

Diana.

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parallelization of LMZ.F3GWR.test in spgwr package

Roger Bivand
Administrator
On Thu, 10 Jan 2019, Diana Gutiérrez wrote:

> Dear all,
>
> I was wondering if it would be possible to run the F3 variability test by
> Leung et al. (2000) (as included in the spgwr package) using parallel, so
> processing times could be reduced when you have a large number of
> observations and parameters. I would appreciate any insights and/or code
> examples if you happen to know of any that can be shared.

Yes, GWR is very demanding for computation. It should only ever be used
for exploring model mis-specification, never for inference or prediction.
So making a bad test on a bad method run faster should not be a priority.

Roger

>
> Thanks in advance,
>
> Diana.
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: [hidden email]
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Roger Bivand
Department of Economics
Norwegian School of Economics
Helleveien 30
N-5045 Bergen, Norway
Reply | Threaded
Open this post in threaded view
|

Re: Parallelization of LMZ.F3GWR.test in spgwr package

Rolf Turner


On 1/13/19 1:06 AM, Roger Bivand wrote:

> ... GWR is very demanding for computation. It should only ever be used
> for exploring model mis-specification, never for inference or prediction.
> So making a bad test on a bad method run faster should not be a priority.

Fortune nomination!

cheers,

Rolf

--
Honorary Research Fellow
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parallelization of LMZ.F3GWR.test in spgwr package

Erin Hodgess
Would there actually be any interest in making GWR parallel, please?  I
enjoy messing with parallel stuff, but don’t want to spend time if it is
not productive.

Thanks for listening!
Sincerely,
Erin

On Sat, Jan 12, 2019 at 1:56 PM Rolf Turner <[hidden email]> wrote:

>
>
> On 1/13/19 1:06 AM, Roger Bivand wrote:
>
> > ... GWR is very demanding for computation. It should only ever be used
> > for exploring model mis-specification, never for inference or prediction.
> > So making a bad test on a bad method run faster should not be a priority.
>
> Fortune nomination!
>
> cheers,
>
> Rolf
>
> --
> Honorary Research Fellow
> Department of Statistics
> University of Auckland
> Phone: +64-9-373-7599 ext. 88276
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Erin Hodgess, PhD
mailto: [hidden email]

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parallelization of LMZ.F3GWR.test in spgwr package

Roger Bivand
Administrator
On Mon, 14 Jan 2019, Erin Hodgess wrote:

> Would there actually be any interest in making GWR parallel, please?  I
> enjoy messing with parallel stuff, but don’t want to spend time if it is
> not productive.

spgwr::gwr() takes a cl= argument to pass a parallel object into the
function, but this only saves time if the fit.points are given as an
argument, and where # fit points is >> # data points.

Not checked for many years, used in: http://hdl.handle.net/11250/163254 
with no obvious easy advantage. Maybe worth revisiting?

Roger


>
> Thanks for listening!
> Sincerely,
> Erin
>
> On Sat, Jan 12, 2019 at 1:56 PM Rolf Turner <[hidden email]> wrote:
>
>>
>>
>> On 1/13/19 1:06 AM, Roger Bivand wrote:
>>
>>> ... GWR is very demanding for computation. It should only ever be used
>>> for exploring model mis-specification, never for inference or prediction.
>>> So making a bad test on a bad method run faster should not be a priority.
>>
>> Fortune nomination!
>>
>> cheers,
>>
>> Rolf
>>
>> --
>> Honorary Research Fellow
>> Department of Statistics
>> University of Auckland
>> Phone: +64-9-373-7599 ext. 88276
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> [hidden email]
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
>
--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: [hidden email]
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Roger Bivand
Department of Economics
Norwegian School of Economics
Helleveien 30
N-5045 Bergen, Norway
Reply | Threaded
Open this post in threaded view
|

Re: Parallelization of LMZ.F3GWR.test in spgwr package

Erin Hodgess
Thanks, I'll check it out.  Awesome!

Sincerely,
Erin


On Mon, Jan 14, 2019 at 12:21 PM Roger Bivand <[hidden email]> wrote:

> On Mon, 14 Jan 2019, Erin Hodgess wrote:
>
> > Would there actually be any interest in making GWR parallel, please?  I
> > enjoy messing with parallel stuff, but don’t want to spend time if it is
> > not productive.
>
> spgwr::gwr() takes a cl= argument to pass a parallel object into the
> function, but this only saves time if the fit.points are given as an
> argument, and where # fit points is >> # data points.
>
> Not checked for many years, used in: http://hdl.handle.net/11250/163254
> with no obvious easy advantage. Maybe worth revisiting?
>
> Roger
>
>
> >
> > Thanks for listening!
> > Sincerely,
> > Erin
> >
> > On Sat, Jan 12, 2019 at 1:56 PM Rolf Turner <[hidden email]>
> wrote:
> >
> >>
> >>
> >> On 1/13/19 1:06 AM, Roger Bivand wrote:
> >>
> >>> ... GWR is very demanding for computation. It should only ever be used
> >>> for exploring model mis-specification, never for inference or
> prediction.
> >>> So making a bad test on a bad method run faster should not be a
> priority.
> >>
> >> Fortune nomination!
> >>
> >> cheers,
> >>
> >> Rolf
> >>
> >> --
> >> Honorary Research Fellow
> >> Department of Statistics
> >> University of Auckland
> >> Phone: +64-9-373-7599 ext. 88276
> >>
> >> _______________________________________________
> >> R-sig-Geo mailing list
> >> [hidden email]
> >> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> >>
> >
>
> --
> Roger Bivand
> Department of Economics, Norwegian School of Economics,
> Helleveien 30, N-5045 Bergen, Norway.
> voice: +47 55 95 93 55; e-mail: [hidden email]
> https://orcid.org/0000-0003-2392-6140
> https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en

--
Erin Hodgess, PhD
mailto: [hidden email]

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo