Description

MarkovSet is a Dictionary that contains keys pointing to WeighBags that contain objects and their probabilities. By parsing in a stream the Set "learns" what element can possibly follow another.¹

Part of MathLib, a diverse library of mathematical functions.

Class Methods

MarkovSet.new(args, updateSeeds: false)

MarkovSet.fill(length, stream)

Inherited class methods

Instance Methods

.read(prevkey, nextkey)

.put(prevkey, nextkeys, weights, precision: 0.01)

.remove(key, nextKey)

.next(prevKey)

.parse(pattern, length: inf)

.asStream

Inherited instance methods

Undocumented instance methods

.addExits(terminators)

.choose

.dict

.embedInStream(inval, repeats: inf)

.embedSpyInStream(pattern, repeats: inf, inval)

.equalize

.get(key)

.makeSeeds

.manipulate(func)

.nodeClass

.normalize

.order

.parseFile(path, delimiter: $ , blockSize: 0, completionFunc)

.parseIO(iostream, delimiter: $ , blockSize: 0, completionFunc, maxLength: 32)

.putAll(args)

.seeds

.seeds = collection

.size

.storeItemsOn(stream, itemsPerLine: 1)

.updateSeeds

.updateSeeds = value

.writeHtml(argPath: "supermarkov", nPages: 5, pagelength: 60, style)

Examples

direct node definition

( m = MarkovSet([ [100, [120, 130]], [120, [100, 1900]], [130, nil], ]); ) 8.do { m.next(100).postln }; 8.do { m.next(120).postln }; 8.do { m.next(130).postln };

training by a stream

m = MarkovSet.fill(100, Pseq([1, 2, 3, 4, Prand([5, 55, 555])], inf).asStream); 8.do { m.next(1).postln }; 8.do { m.next(4).postln }; 8.do { m.next(55).postln };

training bit by bit

m = MarkovSet.new; m.read(\x, \y); m.read(\x, \z); m.read(\z, \abcde); m.read(\y, nil); 8.do { m.next(\x).postln }; 8.do { m.next(\y).postln }; 8.do { m.next(\z).postln }; m.read(\y, \mmmm); 8.do { m.next(\y).postln }; // now it sometimes terminates m.remove(\y, \mmmm); 8.do { m.next(\y).postln };

creating streams

( m = MarkovSet([ [100, [120, 130]], [120, [100, 1900]], [130, [100]], [1900, [1000, 2000, 3000]], ]); ) m.makeSeeds; x = m.asStream; 32.do { x.next.postln }; // spy out a stream to feed it in z = Pseq([100, 200, 300, Prand([1.2, 1.4, 1.5])],inf).asStream; z = Pspy(m, z).asStream; 32.do { x.next.postln }; 15.do { z.next } // add some more elements from z 32.do { x.next.postln }; // event streams ( SynthDef("mark", { arg out, freq, sustain=1, amp=0.1, pan; var env, u=1; env = EnvGen.kr(Env.perc(0.03, sustain, 5), 1, doneAction:2); 3.do { var d; d = exprand(0.01, 1); u = SinOsc.ar(d * 300, u, rrand(0.1, 0.4) * d, 1) }; Out.ar(out, Pan2.ar(SinOsc.ar(u + 1 * freq, 0, amp * env), pan)); }).store; ) ( a = Pbind( \freq, Pshuf([203, 490, 231, 764, 234], inf), \dur, Pshuf([0.5, 0.5, 1.0, 0.25, 0.125, 0.5], inf) ); m = MarkovSet.fill(20, Pevent(a, Event.default).asStream); ) m.dict.keysValuesDo { |key, val| postf("key: %\t\tvalue: %\n", key, val) }; "" // compare: b = Pchain(Pfsm2(m), (instrument: \mark)); c = Pchain(a, (instrument: \mark)); Ppar([b, c]).play;

supermarkov

m = MarkovSet.new; // read in 2 files m.parseFile("Help/More-On-Getting-Help.html".standardizePath); m.parseFile("Help/Language/Intro-to-Objects.html".standardizePath); // write 5 supermarkov html pages m.writeHtml(nPages:5, pagelength:120); // you can also create a string like this: ( x = Pfsm2(m, inf).asStream; 300.do { x.next.post }; )

Authors

Julian Rohrhuber, 2004, 2007.