comp.lang.idl-pvwave archive: archive

Home » Public Forums » archive » majority voting

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: majority voting [message #65167 is a reply to message #65079]

Wed, 11 February 2009 15:20

JD Smith
Messages: 850
Registered: December 1999

Senior Member

You're trying to compute the n-dimensional mode. For the 1D case,
see:

http://www.dfanning.com/code_tips/mode.html

You could of course use SORT_ND and a 2D version of the "find the
longest string of the same number" method on this page. This is very
safe against sparse input distributions -- notice the HISTOGRAM in
SORT_ND is of the sort indices, not the data themselves. It's also
easy to recognize a mode length of 1 as "no mode", and (if you like)
detect multiple modes in the data (aka "tie votes").

However, your problem is a special case, since (I presume) your vote
labels are always low contiguous integers, you don't have to worry
about sparseness. In this case, using HIST_ND directly without
sorting will likely be much faster:

n=n_elements(array)
s=size(array,/DIMENSIONS)
h=hist_nd([lindgen(1,n)/s[1],reform(transpose(array),1,n)],1 )
m=max(h,DIMENSION=2,mode) gt 1
mode=mode/s[0] * m + m - 1

All I'm doing here is forming a 2d space, with the first dimension the
integer ID of the vote (aka column number in your example), and the
second the actual votes cast.

Had you organized your array with the rows as your "vote items" you
could save the transpose. Also note I specifically test for and set
to -1 any mode with frequency of 1 (your "don't care"). I do not
check for multiple modes, but you could do this as well in a
straightforward way. The fact that the dimensions of the 2D histogram
are the same as the input array is incidental, by the way.

JD

Report message to a moderator

[Message index]

		majority voting By: Mort Canty on Wed, 11 February 2009 08:14
		Re: majority voting By: Allan Whiteford on Fri, 13 February 2009 01:31
		Re: majority voting By: David Fanning on Sat, 14 February 2009 08:44
		Re: majority voting By: JD Smith on Thu, 12 February 2009 14:45
		Re: majority voting By: Foldy Lajos on Thu, 12 February 2009 12:47
		Re: majority voting By: David Fanning on Thu, 12 February 2009 12:11
		Re: majority voting By: pgrigis on Thu, 12 February 2009 11:58
		Re: majority voting By: JD Smith on Thu, 12 February 2009 10:49
		Re: majority voting By: Mort Canty on Thu, 12 February 2009 07:45
		Re: majority voting By: JD Smith on Thu, 12 February 2009 07:08
		Re: majority voting By: David Fanning on Thu, 12 February 2009 05:43
		Re: majority voting By: Allan Whiteford on Thu, 12 February 2009 04:17
		Re: majority voting By: Mort Canty on Thu, 12 February 2009 01:15
		Re: majority voting By: JD Smith on Wed, 11 February 2009 15:20
		Re: majority voting By: ben.bighair on Wed, 11 February 2009 14:03
		Re: majority voting By: Allan Whiteford on Mon, 16 February 2009 04:11

Previous Topic:	Re: Reading multiple ASCII files in as 2d arrays and putting them into a 3d array
Next Topic:	Different issue

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Oct 08 16:04:09 PDT 2025

Total time taken to generate the page: 0.00430 seconds