This BSc thesis describes the means of generating music sequences using an agent-based model. As a response to surrounding events, agents (entities) emit certain musical tones. Individual entity is capable of several different responses, which in combination with responses of the remaining entities form musical sequences. Entities aren't aware of the end goal. After every execution of a certain number of steps, their contribution is assessed by a fitness function, which dictates the shape of the desired solution. The score of the individual entity consists of partial assessments of different musical aspects of the produced sequence. It determines the chance of mutation of the responses' parameters of a new generation of an entity. Created directional search algorithm searches for parameter values and sequence of responses that the entities must perform, so that in the case of a set of entities the produced sequences would form a coherent whole.
|
