Thread Subject: angular comparison?

Hi there,
I figured this would be a common problem, so there might be a simple trick already worked out.

How do I cluster variables that represent angles such that a value of 1 degree is deemed 'close' to 359 degrees?

Eg.

angles = rand(1,20)*360;

If I were working in a linear space a simple difference of two angles would give me their difference. But I think that my space is cyclic, so I'm not quite sure how to best have my space 'wrap around' from 360 back to 0, in preparation to do some clustering.

Thanks,
Sven.

"Sven" <sven.holcombe@gmail.deleteme.com> wrote in message <h7rmod$ii4$1@fred.mathworks.com>...
> Hi there,
> I figured this would be a common problem, so there might be a simple trick already worked out.
>
> How do I cluster variables that represent angles such that a value of 1 degree is deemed 'close' to 359 degrees?
>
> Eg.
>
> angles = rand(1,20)*360;
>
> If I were working in a linear space a simple difference of two angles would give me their difference. But I think that my space is cyclic, so I'm not quite sure how to best have my space 'wrap around' from 360 back to 0, in preparation to do some clustering.
>
> Thanks,
> Sven.

To name a few,

You could compare the sin (or cosine) of the angles:
doc sind
doc cosd

Or, you could use the modulus function on the angles directly
doc mod
doc rem

I'm sure there are many other ways alao.