Im having difficulty running multiple markov chains in one patch. Basically i am using a webcam to trigger different markov chains with different midi files in them. I cant seem to workout how to separate the chains so they can each have a different midi file allocated to them.

Apologies for the very basic way this is explained, i am new to PD.

I appreciate any help with this.

Thanks
Jon

have you checked out the markov abstractions in rjlib, they have great help patches.

What you would want to do is save the histograms that can be dumped by those abstraction and reload them according to the midi files you load.

hi,

Thanks for your quick response i appreciate it.

I just looked up what you suggested but unfortunately i don't really understand any of it. Is there maybe a way you could it explain it to me in really basic terms.

Thank you again

Jon

A markov chain is a set of proba, for each note of a given music. The workflow is as follow :

• you play a midi file and record the playing instruction using [m_markovanal]
• when playing is finished you send a [dump( message to the right inlet, and use [u_cocollect] to get an histogram
• you load that histogram in [c_markov] right inlet and set states in the left inlet.

Do you have any code you could share, to work on ?

Hey everyone,

I had a quick question regarding the [c_markov] object. From what I have understood the probabilities of each state happening (transition matrix) is inputed in the right inlet as a series of messages (as can be seen in the help patch). For example, if I had 3 states I would have a 3x3 matrix that can be inputed by a message as follows:

[0 0.6 0.4, 0.3 0.3 0.4, 0.8 0.1 0.1(

What I understand is that the sum of the probabilities of each row must sum up to 1, but seeing the example shown in the help file this is not done like this (and doing some tests this also doesn't seem to be how it works). Can anybody correct me if I'm wrong and point me a bit in the right direction?