| for loops for total newbies [message #54800] |
Fri, 13 July 2007 06:54  |
Anna
Messages: 2
Registered: July 2007
|
Junior Member |
|
|
Dear experts,
[skippable] I'm a physics student currently working on my bachelor
thesis research. To prepare me for the programming aspects, I've been
taught C/C++ by an assistant professor in C.S. who has advised a
gazillion Master students on their thesis, as well as several Ph.D's,
and can spot a syntax error from across the computer lab, and LabView
by a grad student in physics whom we suspect was caught while getting
coffee and forced into teaching the course although he unfortunately
doesn't speak the language in which the manual was written (our mother
tongue), and prefers to keep his comments to "Your program, it does
not work".
We suspect that the philosophy behind these teaching methods is that
we're now supposed to be able to work with all languages by applying
linear combinations of our experiences. However, I am now supposed to
use IDL to analyse my raw data, and I'm stuck. I've tried asking some
astronomy majors but they turned completely white and ran away. [/
skippable]
For my research project, I have to analyse sets of png images (each
set consisting of 1000 210x820 images). I have an IDL program which
can be coerced into working on a good day, but running it over the
data of one day takes a night and a morning, which means that I can't
measure the next morning (never mind what the hardworking sysadmins
think of me). I thought I could combine two things: optimising my
program, and teaching my C-trained, for-loop-loving mind the power of
IDL by asking you to look at two pieces of code:
===================================
cor_row= fltarr(max_cor_shift)
for i = 0, number_of_images-3 do begin
print, i
for j = 0, length-1 do begin
for lag = -10, 10 do begin
cor_row(lag+10)= c_correlate(data(i
+1,*,j),data(i,*,j),-lag)
endfor
cor_array(i,j,*)=cor_row
endfor
endfor
===================================
'data' is an array containing all images, of dimension
data(number_of_images, width, length) = (1000, 210, 820).
'cor_array' is supposed to contain all correlation data at the end,
and has dimenson (number_of_images - 2, lenght, max_cor_shift).
max_cor_shift is a constant and in this case 21.
This is reasonably fast, but I'd like to know whether it can be
written differently, both to make it faster and as a learning
experience...
The real cycle eater is the fitting function:
==========================
crtot= fltarr(sz-1,(size(cor_array))[2],(size(cor_array))[3])
crtot(0:sz-3,*,*)= cor_array(0:sz-3,*,*)
meanarr = fltarr((size(cor_array))[2])
j= fltarr(1000*max_cor_shift)
for i=0,((1000*max_cor_shift)-1) do j(i)=i/1000.
avgz= fltarr(sz-2)
fit = fltarr(sz-2, max_cor_shift*1000)
for i = 0, (size(cor_array))[2]-1 do begin
print, i
for f=0, number_of_images-3 do begin
wait,.0001
maxcr=max(crtot(f,i,*),k)
k2=(k+1+max_cor_shift) mod max_cor_shift
k1=(k-1+ max_cor_shift) mod max_cor_shift
a= crtot(f,i,k2)/2+ crtot(f,i, k1)/2- maxcr
b= crtot(f,i,k2)/2- crtot(f,i, k1)/2
c= maxcr
fit(f,*)= a*(j-k)^2 + b*(j-k) + c
kaas= max(fit(f,*), s)
avgz(f)= s/(float(sz)) -10
endfor
meanarr(i)=mean(avgz)
endfor
===============================
This takes about two hours per data set. I tried averaging earlier,
but that had too much of an impact on the quality of the fits.
If you could please help me, I'd be very, very much obliged!
Thank you,
Anna Baas
|
|
|
|
| Re: for loops for total newbies [message #54877 is a reply to message #54800] |
Mon, 16 July 2007 02:14  |
Paolo Grigis
Messages: 171
Registered: December 2003
|
Senior Member |
|
|
Anna wrote:
> [skip]
>
> ===================================
> cor_row= fltarr(max_cor_shift)
> for i = 0, number_of_images-3 do begin
> print, i
> for j = 0, length-1 do begin
> for lag = -10, 10 do begin
> cor_row(lag+10)= c_correlate(data(i
> +1,*,j),data(i,*,j),-lag)
> endfor
> cor_array(i,j,*)=cor_row
> endfor
> endfor
> ===================================
>
here's another suggestion: c_correlate accepts an array
as lag, so instead of looping over lag from -10 to 10
you could just give it an array [-10,-9,..., 9,10] as
input, thus eliminating the innermost loop.
Ciao,
Paolo
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
