Thread Subject: Unshuffle algorithm

Hi there,
I have an array of numbers made up of two sequences that
have been sorted. I am trying to find an algorithm that can
extract separate arrays made up of these sequences.

For example:
arr1 = sort([1:5 7:0.5:10]);

This first case is straight-forward, as the two sequences
don't overlap. I can get the 1st index of the two sequences
using diff:
indices = find(diff(arr1,2))+2; % +2 because a 2nd order
diff is taken.
sequence_indices = {1:indices(1)-1, indices(1):length(arr1)}

However, this technique only works if the two sequences are
non-overlapping. For example, a more difficult case is:
arr2 = sort([1:5 4.2:10]);

Can anyone point out I should tackle the problem of
extracting the two sequences in this case?

Thanks for any suggestions,
Sven.

"Sven " <sven.holcombe@removethis.gmail.com> wrote in message
<f91dom$3hv$1@fred.mathworks.com>...
> Hi there,
> I have an array of numbers made up of two sequences that
> have been sorted. I am trying to find an algorithm that can
> extract separate arrays made up of these sequences.
>
> For example:
> arr1 = sort([1:5 7:0.5:10]);

If I might ask, what is the source of this problem?
Its exactly the second time I've seen it come up,
the last being roughly 1.5 years ago.

This is, as I see it, a variant of clustering problem.
I can supply a code that does it, using my
consolidator tool as part of the engine. I'd not
seen a reason to post it on the file exchange the
first time it was asked.

Its a bit long to paste in here though. Send me an
e-mail if you want it.

John

"John D'Errico" <woodchips@rochester.rr.com> wrote in
message <f91sg2$3gh$1@fred.mathworks.com>...
> "Sven " <sven.holcombe@removethis.gmail.com> wrote in message
> <f91dom$3hv$1@fred.mathworks.com>...
> > Hi there,
> > I have an array of numbers made up of two sequences that
> > have been sorted. I am trying to find an algorithm that can
> > extract separate arrays made up of these sequences.
> >
> > For example:
> > arr1 = sort([1:5 7:0.5:10]);
>
> If I might ask, what is the source of this problem?
> Its exactly the second time I've seen it come up,
> the last being roughly 1.5 years ago.
>
> This is, as I see it, a variant of clustering problem.
> I can supply a code that does it, using my
> consolidator tool as part of the engine. I'd not
> seen a reason to post it on the file exchange the
> first time it was asked.
>
> Its a bit long to paste in here though. Send me an
> e-mail if you want it.
>
> John
>

Hi John, thanks for the input.
Actually it was for a way to unshuffle a set of dicom images
that had been converted to flat images without metadata. The
filenames were ordered correctly spatially, but this meant
that two separate series had been intertwined because they
overlapped. I'm using past tense here because the original
dicom files have now been tracked down, and I'm sure that I
can now use the metadata to sort things out correctly.
So it's no longer an issue to work out an unshuffling
algorithm, but thanks very much for your interest.

Cheers,
Sven.