How to detect isolated points in a vector

Hi,

I have a vector (see attachment), and I would like to detect isolated points (marked with red cicles in the image). Isolated in that sense, that same values have a certain distance to these. I would need the indices.

Thank you

Chris

[Cris LaPierre]

For others who may be wondering

load outputVector.mat
plot(iidx,'*')

[Dyuman Joshi]

"Isolated in that sense, that same values have a certain distance to these."

Can you elaborate on this? If possible, please provide an example with random set of values.

[CSCh]

Dear Joshi, thanks for your answer. please have a kook to the provided example, in the attachment.

[Dyuman Joshi]

I have seen the image. It does not do a good job of explaining what you mean by "isolated points".

That is why I asked for more information.

Provide a mathematical explaination of the criteria to use to determine which points are isolated points.

Make it easy for us to help you.

Answer 1

[Harald]

Hi,

I believe this code does what you ask for:

load outputVector.mat
N = 500;
isolated = false(size(iidx));
L = length(iidx);
for k = 1:L
    left = max(1,k-N):k-1;
    right = k+1:min(L,k+N);
    isolated(k) = ~any(iidx([left, right]) == iidx(k));
end
result = find(isolated);

plot(iidx, "*")
hold on
plot(result, iidx(isolated), "ro")
hold off

I don't see an obvious way to avoid the loop. If there are concerns about performance, I can think about it a bit more - or perhaps, others have an idea.

Note that it does not identify two of the points that you have circled. If you zoom in, you will see that there are two consecutive identical values there. My understanding is that you did not want to extract these, even though they look isolated on the first glance? If you wanted these, I do not fully understand the question yet.

Best wishes,

Harald

[CSCh]

Hi Harald, thank you so much.