[mindcrime]

Could you have the ML part done as a standalone process and not actually linked into your other project code? That is, process some data "out of band" so to speak, and write it out to a file, which is then used by "your stuff"? Or communicate with the ML part over a socket? Either approach would let you use this code without requiring the other parts of your project to be licensed under the GPL.

Just a thought...

[JacobiX]

Yes I think It is possible to use the library as a standalone process. It will introduce some little overhead. More importantly the code will be harder to debug and maintain.