markov models and max msp ... ?

    anamaria
    Dec 20 2010 | 9:55 am
    Hi,
    Have anyone implemented a markov model in max msp that can learn a specific play style according to user input?
    I am interested in making a machine in max that learns' the user's play style and can come up with a similar play style as a result.
    I am not interested in having a bank of possible play styles to tap into. I want to generate the playstyle according to user input only.
    Does anyone have a clue where to find more information about this and how to implement such a thing in max msp? (I am very inspired of François Pachets work "The Continuator").
    Cheers,
    ana
    • mzed
      Dec 20 2010 | 5:41 pm
      I might start by reading some of Robert Rowe's books. Specifically, Interactive Music Systems and Machine Musicianship. See: http://homepages.nyu.edu/~rr6/ Also, look into George E. Lewis' piece called Voyager. The approach many have taken has been to break "style" down in to a number of parameters that are easier to quantify. "modality", "tempo", "density", &c. I would start simple and build up complexity until the result started to become satisfying.
      mz
    • ADHD101
      Dec 25 2010 | 4:24 pm
      Use presets.
    • yas
      Dec 25 2010 | 10:43 pm
      i saw a tutorial on markov chains in max:
      http://algorithmiccomposer.blogspot.com/2010/05/algorithmic-composition-markov-chains.html
      looks like it's intended for reading midi files, but it'd be pretty simple to adapt it for live input