| Re: array convol optimizationv slow now [message #67282] |
Mon, 13 July 2009 16:53  |
![Chris[6] is currently offline](theme/default/images/xoffline.png.pagespeed.ic.XRkd1fkXye.png "Chris[6] is currently offline")
Chris[6]
Messages: 84
Registered: July 2008
|
Member |
|
|
On Jul 13, 1:26 pm, David Fanning <n...@dfanning.com> wrote:
> newerjazz writes:
>> each particle is a single molecule of protein; we need to visualize
>> all the proteins to get meaningful interpretation.
>
> How does that work? If I put 100 million dots on my screen, things
> get a little blurry. (Of course, my eyes aren't what they used to
> be, admittedly.)
>
>> do you have any suggestions of getting the x,y locations on the image
>> w/ an array operation instead of a for loop?
>
> No, I can't think of any way to do this an an array. :-(
>
> Cheers,
>
> David
> --
> David Fanning, Ph.D.
> Coyote's Guide to IDL Programming (www.dfanning.com)
> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
I've written a routine called densitymap which makes smoothed surface
density maps based on the position of 2d sources. You might want to
check it out (it has been moderately debugged)
http://www.ifa.hawaii.edu/users/beaumont/code/
It uses a different smoothing scheme (it uses the distance to each
pixel's n-th nearest neighbor to estimate the surface density).
If you use it, you should probably download the IDL astronomy user's
library, as well as
http://www.ifa.hawaii.edu/~beaumont/code/beaumont_library.ta r
to avoid dependency issues.
Otherwise, you could use hist_2d to pixellate the locations of each
point, and then convolve the histogram with a gaussian
Chris
|
|
|
|
|
|
| Re: array convol optimizationv slow now [message #67285 is a reply to message #67284] |
Mon, 13 July 2009 16:10  |
newerjazz
Messages: 5
Registered: July 2009
|
Junior Member |
|
|
Thanks David,
each particle is a single molecule of protein; we need to visualize
all the proteins to get meaningful interpretation.
i can look at 1/10 of the particles by reducing the field of view but
it's still slow.
do you have any suggestions of getting the x,y locations on the image
w/ an array operation instead of a for loop?
Thanks!
newerjazz
On Jul 13, 3:57 pm, David Fanning <n...@dfanning.com> wrote:
> newerjazz writes:
>> I have an array of [2,100,000,000] corresponding to x, y and
>> 100,000,000 particles. I need to render these particles as gasussian
>> spots at x, y locations
>
>> Currently, I run a for loop through each particle and render each
>> particle in this for loop; it takes forever.
>
> Not to mention it probably takes a display screen
> the size of a city block to render the mess in
> the kind of resolution that allows you to actually
> make something of it.
>
> Have you tried plotting a few hundred (maybe even
> a few thousand) random points to see if you can draw
> the same conclusions you were drawing by plotting all
> the points?
>
> Cheers,
>
> David
>
> --
> David Fanning, Ph.D.
> Coyote's Guide to IDL Programming (www.dfanning.com)
> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: array convol optimizationv slow now [message #67286 is a reply to message #67285] |
Mon, 13 July 2009 15:57 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
newerjazz writes:
> I have an array of [2,100,000,000] corresponding to x, y and
> 100,000,000 particles. I need to render these particles as gasussian
> spots at x, y locations
>
> Currently, I run a for loop through each particle and render each
> particle in this for loop; it takes forever.
Not to mention it probably takes a display screen
the size of a city block to render the mess in
the kind of resolution that allows you to actually
make something of it.
Have you tried plotting a few hundred (maybe even
a few thousand) random points to see if you can draw
the same conclusions you were drawing by plotting all
the points?
Cheers,
David
--
David Fanning, Ph.D.
Coyote's Guide to IDL Programming (www.dfanning.com)
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: array convol optimizationv slow now [message #67370 is a reply to message #67282] |
Tue, 14 July 2009 08:31 |
newerjazz
Messages: 5
Registered: July 2009
|
Junior Member |
|
|
Thanks Chris,
I will try hist_2d to see if it helps.
Dave, we know the x,y position down to subpixel resolution; so instead
of say plotting particles on a 512x512 grid; we plot it on 5120x5120
grid (if our resolution is 1/10 of a pixel). In practice our
resolution is down to 1/100th of a pixel. Depending on how many
particles are packed into a pixel or a voxel since I actually have z
resolution too; the image can look either sharp or blur.
cheers,
newerjazz
On Jul 13, 4:53 pm, Chris <beaum...@ifa.hawaii.edu> wrote:
> On Jul 13, 1:26 pm, David Fanning <n...@dfanning.com> wrote:
>
>
>
>> newerjazz writes:
>>> each particle is a single molecule of protein; we need to visualize
>>> all the proteins to get meaningful interpretation.
>
>> How does that work? If I put 100 million dots on my screen, things
>> get a little blurry. (Of course, my eyes aren't what they used to
>> be, admittedly.)
>
>>> do you have any suggestions of getting the x,y locations on the image
>>> w/ an array operation instead of a for loop?
>
>> No, I can't think of any way to do this an an array. :-(
>
>> Cheers,
>
>> David
>> --
>> David Fanning, Ph.D.
>> Coyote's Guide to IDL Programming (www.dfanning.com)
>> Sepore ma de ni thui. ("Perhaps thou speakest truth.")
>
> I've written a routine called densitymap which makes smoothed surface
> density maps based on the position of 2d sources. You might want to
> check it out (it has been moderately debugged)
>
> http://www.ifa.hawaii.edu/users/beaumont/code/
>
> It uses a different smoothing scheme (it uses the distance to each
> pixel's n-th nearest neighbor to estimate the surface density).
>
> If you use it, you should probably download the IDL astronomy user's
> library, as well ashttp://www.ifa.hawaii.edu/~beaumont/code/beaumont_library. tar
> to avoid dependency issues.
>
> Otherwise, you could use hist_2d to pixellate the locations of each
> point, and then convolve the histogram with a gaussian
>
> Chris
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
