Making Pairs of Indices in Matrix- Pairwise Function

Hello,

I want to make a pairwise function in matlab simialar to how it is done in python. Or is there one already in matlab?

https://more-itertools.readthedocs.io/en/stable/api.html#more_itertools.pairwise

Overall, I am trying to return a list of pairs of indices in my matrix and then take the distance between each pair.

Any input would be greatly appreciated.

Thank you!

Answer 1

[Matt J]

It depends what Matlab data-type you consider to be the equivalent of a Python tuple. I tend to think of cell arrays as the equivalent of a tuple because cells can store sequences of mixed types, but matrix form is generally the form in Matlab that will support most mathematical operations you might want to do. Below, I generate a set of pairs in both matrix and cell array form.

n=4;
pairMatrix=[0:n-1;1:n].';
pairCell=num2cell(pairMatrix,2)

Overall, I am trying to return a list of pairs of indices in my matrix and then take the distance between each pair.

You mean, the Euclidean distances when each pair is viewed as a 2D coordinate? If you have the Statistics Toolbox, then this can be done using pdist2,

>> pdist2(pairMatrix,pairMatrix)

ans =

         0    1.4142    2.8284    4.2426
    1.4142         0    1.4142    2.8284
    2.8284    1.4142         0    1.4142
    4.2426    2.8284    1.4142         0

[Andrea Lucia Alfonso]

What I am trying to get out is more like this:

If you had a list of numbers [1,2,3,4] , the pairwise function would return [(1,2),(2,3),(3,4)].

My matrix is very large and binary but I would put this parwise function to each individual row first.

For example,

row = [0,0,1,0,0,1,0,0,0,0,1] so the non zero values are at 3,6, and 11

i need something that would output: pairwise = [(3,6),(6,11)] so i could take the distance between each pair in a pairwise and then replace the zeros in between the nonzeros with their respective distance between them

To get something like this

row = [0,0,1,3,3,1,5,5,5,5,1]

[Matt J]

Here's one way to do that:

row = double([0,0,1,0,0,1,0,0,0,0,1]);

loc=find(row); %locations of 1's
D=diff(loc);  %differences

j=0;
for i=loc(1):loc(end)
    
   if row(i)==1, 
       j=j+1; 
   else
      row(i)=D(j); 
   end 
 
end

row,

Here's another, without loops:

row = double([0,0,1,0,0,1,0,0,0,0,1]);

loc=find(row); %locations of 1's
D=diff(loc);  %differences

c=cumsum(row); %cumulative sum along row

mask=c>=1 & c<=numel(D) & ~row; %binary mask of the 0's that are sandwiched by 1's

row(mask)=D(c(mask))

[Andrea Lucia Alfonso]

Thank you so much! This definitley helped!

[Matt J]

You're quite welcome, but please Accept-click the answer to signify that you are satisfied with the solution.