TriScatteredInterp on scattered dataset with duplicates

I am facing a problem that is driving me crazy and I hope that someone will be able to help me...

According to the MATLAB documentation, applying the TriScatteredInterp function on a scattered dataset with duplicates will implicitly remove the duplicates and average the associated data values. Now consider the simple test code

clear all;

% Create x- and y-coordinates of three random points in the 2D plane
X = rand(3,2);

% Duplicate all three points
X = [X;X];

% Assign value 100 to original points and value 0 to duplicates
V = [100*ones(3,1);zeros(3,1)];

% Merge points and values into one matrix
A = [X V];

% ! Critical line !
% A = sortrows(A);

% Construct interpolant
% Duplicates should be removed and associated values should be averaged to 50
F = TriScatteredInterp(A(:,1:2),A(:,3));

% Output values
F.V

In this form, the output is

ans =
    50
    50
    50

as expected. However, upon uncommenting of the critical code line with the sortrows command, the output becomes

ans =
    50
   100
    50

and something is wrong. I believe that the order of the scattered data (and hence the critical code line) should not play any role, should it? What's going on here? I am using MATLAB R2010a.

Any help is appreciated.

Rui

Answer 1

[Jonas]

Er... I obtain the 50-50-50 in both cases...

[Jonas]

But I have this warning, in fact: "Warning: Duplicate data points have been detected and removed - corresponding values have been averaged. " In both cases. I have the 2012a version. Maybe this bug has been corrected between your Matlab version and my one?

Answer 2

[Rui]

Hi Jonas,

Many thanks for having a look at my issue. I appreciate it.

It's strange though that you couldn't reproduce the problem. I get the same warning message, which would be fine if the averaging was done correctly (but in my case it's not). It seems that this is indeed related to the different MATLAB versions. Unfortunately I don't have access to newer versions of MATLAB, so I can't confirm myself whether that's true or not.

Anyway, I will then remove the duplicates and carry out the averaging manually. Any further comments are still appreciated.

Rui