Calculate distances between data points

Hi

I have 6 separated groups(as g1,g2,g3,..., g6) and each of them have some spatial data points as (x,y,z,k) where x,y,z determine position of each data point and k is its value, for example Temperature value at that point.

Note that the number of data points in each group is not equal.

I want to calculate distances between data points in each group and other data points in other groups.

Would you please guide me to solve my problem?

Thanks a lot.

Mani

[Guillaume]

So do you want the distance between all of the points in one group and all of the points in another group, that is size(g1, 1) * size(g2, 1) distances, or something else?

[Mani Ahmadian]

Yes, I do. I want it exactly.

Answer 1

[Guillaume]

Possibly, this is what you want:

g1 = randi([-100 100], 5, 4); %example data, a 5*4 matrix 
g2 = randi([-100 100], 3, 4); %example data, a 3*4 matrix

[idx1, idx2] = ndgrid(1:size(g1, 1), 1:size(g2, 1)); %row indices to calculate distance between
distance = arrayfun(@(row1, row2) norm(g2(row2, 1:3) - g1(row1, 1:3)), idx1, idx2)

Each row in the output matrix correspond to a row of g1, and each column to a row of g2. For example, the value at row 3, column 2 is the distance between row 3 of g1 and row 2 of g2.

[Mani Ahmadian]

I completed a script to solve my problem as bellow:

clc
close all
clear all

% g1=[1 1 1 1; 2 2 2 2; 3 3 3 3; 4 4 4 4; 5 5 5 5];
% g2=[6 6 6 6; 7 7 7 7; 8 8 8 8; 9 9 9 9];
% g3=[10 10 10 10; 11 11 11 11];

tic

g1=randi(1000,1000,4);
g1(:,1)=2;
g1(:,2)=2;

g2=randi(1000,1000,4);
g2(:,1)=5;
g2(:,2)=5;

g3=randi(1000,1000,4);
g3(:,1)=13;
g3(:,2)=13;

rows=[size(g1,1),size(g2,1),size(g3,1)];
cols=[size(g1,2),size(g2,2),size(g3,2)];
MaxRows=max(rows);
MaxCols=max(cols);

n=3;
MyNullValue=-123456.123456;

MyTotal=ones(MaxRows,4,n)*MyNullValue;

MyTotal(1:size(g1,1),:,1)=g1;
MyTotal(1:size(g2,1),:,2)=g2;
MyTotal(1:size(g3,1),:,3)=g3;

MyTotal(MyTotal==MyNullValue)=0;

MyDist=ones(MaxRows,MaxRows,n^2)*MyNullValue;
kk=0;

for ii=1:n

      for jj=1:n

              Mat1=MyTotal(:,:,ii);

              Mat1(all(~Mat1,2), : ) = [];
              Mat1(:,all(~Mat1,1)) = [];

              Mat2=MyTotal(:,:,jj);
              Mat2(all(~Mat2,2), : ) = [];
              Mat2(:,all(~Mat2,1)) = [];

              kk=kk+1;
              MyDist(:,:,kk)=distances(Mat1,Mat2,MaxRows);

      end
  end

MyDist(MyDist==0)=inf;
MinDist=min(min(min(MyDist)))

MyDist(MyDist==inf)=0;
MaxDist=max(max(max(MyDist)))

beep

toc

%'%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5

function [distance]=distances(g1,g2,MaxRows)

[idx1, idx2] = ndgrid(1:size(g1, 1), 1:size(g2, 1)); %row indices to calculate distance between
distance = arrayfun(@(row1, row2) norm(g2(row2, 1:3) - g1(row1, 1:3)), idx1, idx2);

MyRows=size(distance,1);
MyCols=size(distance,2);

if MyRows<MaxRows
    distance=vertcat(distance,zeros(MaxRows-MyRows,MyCols));

end

if MyCols<MaxRows
    distance=horzcat(distance,zeros(size(distance,1),MaxRows-MyCols));

end

end

I run my script just for 3 groups of data points instead of 6 groups as above. It takes long about 5 minutes. How can I change my script to become faster than now?

Thanks a lot

Mani

[matt dash]

Look on the file exchange for the file named "IPDM" http://www.mathworks.com/matlabcentral/fileexchange/18937-ipdm--inter-point-distance-matrix ... it does what you need and it is very fast: