Self-Organizing Map for Automated Quality Assessment in Algorithmic Music Computation
This research work is a part of a larger study that was done to generate music algorithmic by process of Differential Evolution. In this paper, Self Organizing Maps (SOM) are employed as a Fitness Function and discriminator across emotional states, guiding the evolutionary process towards the targeted emotional direction.
In this work, we explore the challenges of modeling a Fitness Function for our specific problem statement, particularly given the scarcity of datasets available for generating music in our niche. While traditional algorithms such as Fuzzy C Means and DBSCAN fail to achieve even rudimentary levels of success, SOM effectively assess composition quality based on defined metrics and generate continuous quality scores that capture nuanced improvements in the music produced.
Resources/References
- Link to the complete paper Self-Organizing Map for Automated Quality Assessment in Algorithmic Music Computation
- The slides used to present the paper at ICETCI 2024: Slides
- This paper talks about the Differential Evolution process: Emotion Aligned Music Composition from Sound Fundamentals Using Differential Evolution