| Can these nested loops be made to run faster? [message #49750] |
Wed, 16 August 2006 02:39 |
humphreymurray
Messages: 8
Registered: August 2006
|
Junior Member |
|
|
Hi,
I am trying to optimize this code. Can anybody suggest how I can make
it run faster? This code is already much more efficient then when I
first wrote it. A description of what I am trying to do can be found
on my previous post:
http://groups.google.com/group/comp.lang.idl-pvwave/browse_t hread/thread/fa5c9a7dca215392/188ba05a3a095f6c#188ba05a3a095 f6c
In short, training_data is a 2d array, which contains sample of pixel
values, and training_classes contains an integer representing what
classes these pixels belong to. testing_data is a 3d array, of which I
am trying to classify. This is similar to my previous post, except I
am now working with pixels in a 2d image.
Before you ask why I have nested loops, the other loop is processing
each row of the image at a time, so that I don't run out of memory. If
I do the operations of all at once, it crashes with a memory error. If
the inner loop can be removed, or some operations removed from it, I
believe this function will speed up immensely.
Cheers
Humphrey Murray
Student of Bachelor of Computing Honours
University of Tasmania, Australia
; Loop through for each row of the image
for y=0, y_size-1 do begin
training_duplicates = REBIN(TRANSPOSE(training_data), num_attributes,
num_training_elements, x_size)
testing_duplicates =
rebin(transpose(testing_data[*,y,*]),num_attributes,num_trai ning_elements,x_size)
euclidean = sqrt(TOTAL((training_duplicates-testing_duplicates)^2, 1))
; calculate the distances for each training item
for x = long(0), x_size - 1 do begin
; Calculate the distances and sort the indexs of these
sorted_indexs = sort(euclidean[*,x])
; Create an array that contains the classes of the items with the k
k_closest_classes = training_classes[sorted_indexs[0:k-1]]
; Calculate the Mode
distfreq = Histogram(k_closest_classes, MIN=Min(k_closest_classes)) ;
Calculate the distribution frequency
maxfreq = Max(distfreq) ; Find the maximum of the
frequency distribution.
mode = Where(distfreq EQ maxfreq, count) + Min(k_closest_classes) ;
Find the mode.
; Store the mode (classes with the highest frequency)
result[x,y] = mode[0]
endfor
endfor
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
