Serendipity in Melodic Self-organising Fitness

Created by W.Langdon from gp-bibliography.bib Revision:1.8129

@InProceedings{Loughran:2018:AISB,

• author = "Roisin Loughran and Michael O'Neill",
• title = "Serendipity in Melodic Self-organising Fitness",
• booktitle = "Symposium on Cybernetic Serendipity Reimagined",
• year = "2018",
• editor = "Joseph Corneli and Colin Johnson and Anna Jordanous and Christian Guckelsberger",
• pages = "13--20",
• address = "Liverpool, UK",
• month = "6 " # apr,
• organisation = "Artificial Intelligence and Simulation of Behaviour",
• keywords = "genetic algorithms, genetic programming, Grammatical Evolution, ES in Melody Generation, Serendipitous evolution, Critic-based Fitness",
• URL = "http://aisb2018.csc.liv.ac.uk/symposia.html#S7",
• URL = "http://aisb2018.csc.liv.ac.uk/PROCEEDINGS%20AISB2018/Cybernetic%20Serendipity%20Reimagined%20-%20AISB2018.pdf",
• size = "8 pages",
• abstract = "Employing Evolutionary Strategies (ES) for subjective tasks such as melody writing causes an immediate problem in determining what to use as a fitness measure. By predefining a measure based on genre, musical rules or human opinion, as has been done in previous studies, we may be prematurely limiting the possibilities obtainable by the system, rendering serendipitous discovery impossible. In this paper, we discuss the development of a system that generates its own self-adaptive fitness measure in response to a corpus of evolved melodies. The system dynamically creates new fitness measures, or Critics, in response to new melodies in a cyclical manner with minimum human intervention. Thus it is a closed loop feedback system that develops its own fitness function through a response to its environment. We propose that the development of such a system could lead to more autonomous creativity and that the use of dynamically changing Critics and melodies could encourage the emergence of serendipitous discovery.",
• notes = "Evolving the Critic. Second Order Cybernetic system. http://aisb2018.csc.liv.ac.uk/",

}

Genetic Programming entries for Roisin Loughran Michael O'Neill

Citations