Subsequent generations evolve from the current generation when the geneseq object receives the message 'evolve' in its left inlet. Generation 0 is a random seeded population when the Max object is created. The random score is used so that for some generations, the first is chose and in other generations the second is chosen if they are the two lowest by fitness score alone. For example, given the following two individuals:Д«oth would have a fitness score of 1 using the default target melody above since each is off by only one step (the 8 in each sequence). The purpose of the random score is to add some variation to the sequencer output when choosing a best candidate individual at a particular point in time. Second, they have a random score as a float generated between 0 and 1. A lower score is more fit and closer to the target. First, they have an integer fitness score based on how many steps they have in common with the current target sequence. This individual sequence will be sent out the first/left outlet of the object, along with its fitness score sent out the second outlet. Individuals are scored so that the best melody can be chosen for a given generation. For example, a gene value of 5 is intended to be the 5th scale degree. Numbers 1-8 are scale degrees root/tonic (1) up to the note one octave above (8). 0 is intended to represent a skipped sequence step, or silence. The valid genes that can make up an individual are the integers 0-8 inclusive. The initial population of individuals is seeded from random combinations of genes. It is a melody generated from the list of valid genes. Genes (i.e., the nucleotides): integers 0-8ДЄn individual is a member of the population.Evolution is a process that happens by sending the Max object an 'evolve' message.ДЄt present, the following fixed properties are in place: When at least one robot produces the target melody, the genetic algorithm has converged and ceases to evolve. The robot with a melody most similar to the target is chosen for each evolution/generation for the population as the current sequence. The robots are trying to resolve to a target melody. Think of a population of robots 'improvising' by each generating different random melodies. Evolution is based on a population of individuals. OverviewДЄ genetic algorithm 'evolves' towards a target state, in this case a target melody specified as scale degrees. It was inspired by the Generative sequencers thread from the lines community. This work is an attempt to build a melody sequencer based on a genetic algorithm.

It has been known to crash Max and M4L/Ableton Live when copying an instance of the object. This Max external is a work in progress and is currently buggy. See the Max SDK from Cycling 74's website. You can build this object if you are comfortable building a Max external on your local system. A genetic algorithm melody generator/sequencer for Max/MSP.

Max msp sequencer