Prev: element-by-element multiplication
Next: Error while trying to use cwt.m in wavelet toolbox 2.2
From: Natalie Sin Hwee on 19 Jun 2010 14:24
Dear Mathwork users

I have a 1000x1000 matrix with values between 0-1. How can i find the non-zero minimum from each column?

I've tried using min but that gives me all zeros

Thanks
Natalie^^
From: Matt Fig on 19 Jun 2010 14:34
One approach:


% Example data:
A = round(rand(6,6)*3)

% Get the non-zero mins
A(~A) = nan;
col_mins = min(A)
From: Natalie Sin Hwee on 19 Jun 2010 14:58
"Matt Fig" <spamanon(a)yahoo.com> wrote in message <hvj2it$irj$1(a)fred.mathworks.com>...
> One approach:
>
>
> % Example data:
> A = round(rand(6,6)*3)
>
> % Get the non-zero mins
> A(~A) = nan;
> col_mins = min(A)

Dear Matt,

that worked perfectly!! just what i wanted!!

Thank you!!
Natalie
From: Matt J on 19 Jun 2010 15:27

> % Get the non-zero mins
> A(~A) = nan;
> col_mins = min(A)

I've been advised in past threads that NaN processing is a lot slower than Infs. It would be advisable to use

A(~A) = inf;

instead.
From: Matt Fig on 19 Jun 2010 17:36
Wow, you weren't joking Matt! As far as the writing is concerned, there is not much difference. But for MATLAB to deal with the array, BIG DIFFERENCE.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [] = nan_inf()
% Is dealing with infs faster than writing nans?
t = 0;

for ii = 1:100
A = round(rand(1000));
tic
A(~A) = nan;
colmins = min(A);
t = t + toc;
end

fprintf('Time for the nans is: %.3f seconds.\n',t)
t = 0;

for ii = 1:100
A = round(rand(1000));
tic
A(~A) = inf;
colmins = min(A);
t = t + toc;
end

fprintf('Time for the infs is: %.3f seconds.\n',t)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%





>> nan_inf
Time for the nans is: 16.497 seconds.
Time for the infs is: 4.969 seconds.
>> nan_inf
Time for the nans is: 16.300 seconds.
Time for the infs is: 4.951 seconds.
>>

On winvista 32 using 2007b and an old laptop.
 | 
Pages: 1
Prev: element-by-element multiplication
Next: Error while trying to use cwt.m in wavelet toolbox 2.2