GenoMus: Towards artificial creativity through metaprogramming of musical genotypes

Resumen

This doctoral thesis introduces GenoMus, a generative and bioinspired computational model for the development of artificial musical creativity based on metaprogramming. It is designed for both autonomous use and human-machine collaboration. The model comprises a representation system, a library of generative and auxiliary functions, and an interactive interface for musical experimentation. The representation system defines the genotype as the functional tree of underlying procedures in a musical piece, and the phenotype as the musical outcome of these processes. Both elements interconnect to form a bidirectional generative grammar that serves as both a programming language and an abstract numerical representation. Its design is optimized to streamline one-dimensional encoding, metaprogramming, and seamless integration with any machine learning model utilizing numerical vectors. GenoMus has been developed through a theoretical-practical methodology, undergoing testing in musical creation projects. The model’s design has been shaped by these artistic experiences, evolving iteratively through a cycle of conceptual framework review, algorithm refinement, and practical application in musical compositions, presented as part of this research.