Re: [sc-users] Similarity Clumping

Well, my thinking is that once the threshold is exceeded, a new "first" value would need to be established for the next clump. Also, the array would need to be sorted first. Maybe I'm still missing a scenario.

I believe this would do what I want, but I'm really not sure how to execute it.

Also, wondering if you've had any issues with KMeans quark.. I got the following:

UnitTesting reports an incompatibility with this SuperCollider version or with other already installed quarks.

I removed all other Quarks, without much luck.

On Sat, Nov 30, 2019 at 8:23 PM <jamshark70@xxxxxxxxx> wrote:

On Sun, Dec 1, 2019 at 5:36 AM <jables.deutsch@xxxxxxxxx> wrote:
>
> Ah, that's an interesting approach.. thanks!
> I'm still a little unsure about one last thing, though:
>
> If things are only compared to the previous value, this could produce very long sequences of correlated numbers..
>
> Again, it might be better for me to illustrate:
>
> a = [0.3, 0.31, 0.32, 0.33]
> and the threshold is at 0.1...
> The difference between 0.3 and 0.31 is within the threshold, but then the .separate method compares 0.31 to 0.32. Is there a way to make the method only compare to the first input value?

Hm, best to try your algorithm on paper, with a lot of different
scenarios. I have a feeling "compare to the first input value" may
turn out to be not quite what you want.

You might try the KMeans quark. That's based on a statistical
technique to form clusters from an unordered dataset.

hjh

_______________________________________________
sc-users mailing list

info (subscription, etc.): http://www.birmingham.ac.uk/facilities/ea-studios/research/supercollider/mailinglist.aspx
archive: https://listarc.bham.ac.uk/marchives/sc-users/
search: https://listarc.bham.ac.uk/lists/sc-users/search/