Kolmogorov complexity

Here's an interesting set of papers:
http://homepages.cwi.nl/~paulv/kolmcompl.html

Among which
R. Cilibrasi, R. de Wolf, P. Vitanyi.
Algorithmic clustering of music
http://homepages.cwi.nl/~paulv/papers/music.pdf

The idea is to use a compression method to find
similarities in strings (which can represent anything)
and automatically make classification trees without
having to select features to base them upon.

I also had fun with "The miraculous universal distribution".

Software to try this out yourself is here:
http://complearn.sourceforge.net/

It's no good trying this on the scale archive however, it
doesn't work very good on small files and there are too many
files also; they must be compared pairwise.

But the results can be spectacular. Vitanyi found to his
own surprise the relationships of the SARS virus correctly
months before virologists did after laboratory experiments.

Manuel

> Here's an interesting set of papers:
> http://homepages.cwi.nl/~paulv/kolmcompl.html

Yeah, Vitanyi's stuff is cool.

> Among which
> R. Cilibrasi, R. de Wolf, P. Vitanyi.
> Algorithmic clustering of music
> http://homepages.cwi.nl/~paulv/papers/music.pdf
>
> The idea is to use a compression method to find
> similarities in strings (which can represent anything)
> and automatically make classification trees without
> having to select features to base them upon.

They used it for identifying authorship of old
documents also.

> I also had fun with "The miraculous universal distribution".

Hadn't seen that. I'll check it out...

-Carl