Squish is a classifier based on probabilistic finite state models.  It
effectively implements much of the DMC compression algorithm.  In essence,
it creates several different compression models, one for each classification
"bucket".  To train an input into a bucket, you simply traverse the model,
and increment the counter for each transition, cloning states as you would
in DMC.  To classify an input, you take the sum of the logs of the
probabilities of each input bit as the finite state model is traversed.
Whichever bucket produces the highest numerical result is selected as the
classification for that input.  This is effectively equivalent to multiplying
out all of the probabilities, but can be done much more efficiently.

Squish is written in OCaml and is still experimental.