How to use indices returned from [Y,I] = min(X) when X is a 2D matrix?

I'm sure there is an answer here but I'm just not finding the right keywords...

I have a large matrix X, and I want the minimum values from each column. min(X) does exactly what I want.

But how would I use the indices returned in "I"? X(I) does not return the minimum values from min(X). Is there notation to address X columnwise, so that I would get back a row vector that matches min(X)?

Answer 1

[Jan]

The replied indices are realted to each column. So you need sub2ind to get the linear index to obtain the minimal values:

x = rand(3, 3);
[v, ind] = min(x);
w = x(sub2ind(size(x), ind, 1:size(x, 2))));

[Gabe]

Thanks! sub2ind was just what I needed.

[Titus Edelhofer]

Hi Gabe,

fine, please go ahead then and mark Jan's answer as "the" answer and the question answered ...

Thanks,

Titus

Answer 2

[Stephen23]

Although the documentation is not very clear on this, the function min returns indices along the dimension that it is working along. So when you sum along each column, the indices are for the position in that column. You can use sub2ind to convert these indices into linear indices:

>> X = randi(9,3,3)
X =
     8     9     7
     6     3     4
     5     7     6 
>> [Y,I] = min(X,[],1)
Y =
     5     3     4
I =
     3     2     2
>> Z = sub2ind(size(X),I,1:size(X,2))
Z =
     3     5     8
>> X(Z)
ans =
     5     3     4

[Gabe]

Perfect, thanks a lot