Maps was the result of a research project classifying 600 electronic songs on a self-organizing map. As my first formal research project at CalArts (2008), I wrote Maps under the tutelage of Ajay Kapur and in collaboration with fellow students Owen Vallis, Jordan Hochenbaum, and Jim Murphy. We were motivated to bring touch-based interaction to the browsing of large music collections and to explore the characteristics of electronic dance music on the feature level. In 2009, it won the best poster award at the International Conference for Music Information Retrieval (ISMIR09).

In looking at methods to reduce the dimensionality of a feature set for proper visualization, we were drawn to Kohonen self-organizing maps because of their effective 2D representations. A 2D space was particularly attractive to us since it made touch-based interactions fairly straightforward to both implement and use. The map is sorted via a computationally-efficient k-nearest-neighbor algorithm and then mapped to an RGB vector. While feature extraction and song annotation are done offline by separate software we developed for this project (written in the ChucK language), the map automatically re-sorts if new songs have been added to the library.