This accessible page is a derivative of https://www.researchcatalogue.net/view/2938321/3620189 which it is meant to support and not replace.
Video description: A video showing the creation of a graph based on existing patterns that use growth and preferential attachment. A graph is ‘produced’ while playing and can be traversed at a later stage. The live generation of graphs and resulting sound sequences are the prominent features here and should be clearly audible.
Click on https://www.researchcatalogue.net/view/2938321/3620189#tool-3620399 to watch the video.
Preferential Attachment
Network setups allow for experimenting with relationships between entities and how one thing leads to another. The process is guided by how the connected points activate, focusing on how fast and in what way they change. Having rethought and converted the material sets to relational representations, I implemented methods of going through them. Although it seemed promising in the beginning, implementing arbitrary walks through a predefined graph was not very successful and some features of the data set felt missing. I then started to think of them rather as growth models, where new nodes are added to the network over time instead of initially. Traversing the network not only exposed relationality as music but also transformed the understanding of the network’s construction into a musical experience. This also fitted a more live-coding attitude where one can type, evaluate code parts, and gradually build up the network as the creative process takes place.
My starting point was to implement a classic network algorithm, the Barabási-Albert model. It is designed to capture the growth and preferential attachment mechanisms observed in many real-world networks, where a few nodes accumulate a disproportionately large number of connections while most nodes have only a few connections (Barabási and Albert 1999). Instead of making the graphs move towards equality, they would instead boost certain nodes and create clusters within the graph. Playing around with variations of the basic model, I implemented a few known variations such as the Krapivsky and Redner Model, where rather than remaining forever part of a network once added, it introduces the possibility of edge deletion, in addition to the creation of new edges and nodes (Krapivsky and Redner 2003). Therefore, relations change during the growth process and every step during the growth presents a unique state of the network.
Since these ideas were created purely within the audio programming language SuperCollider, I started to further articulate them through sound. For example, introducing generated sound or silences when nodes are activated, to transform the graph paths through custom audio processing or layer several paths at once, forming polyphonic textures. My goal became to sonify the time-varying aspects of the network model as much as the nodes themselves would become musical material.
Holding Pattern, Protean
Guided by low, repeating bass patterns, ‘Protean’ contains the largest set of accompanying material. As the graph unfolds, detailed singular, noise-textures follow while some of the underlying rhythmic sequences are kept throughout. Most sounds somehow repeat and reappear which is rather contrasting compared to much of the other music there. The work is also predominantly textural where the development of the noise-textures receives an important focus.
Audio description: Sound file of Protean. Guided by low, repeating bass patterns, Protean contains the largest set of accompanying material. As the graph unfolds, detailed singular, noise-textures follow while some of the underlying rhythmic sequences are kept throughout. Most sounds somehow repeat and reappear, in contrast to much of the other music there. The work is also predominantly textural, where the development of the noise-textures receives an important focus.
Click on https://www.researchcatalogue.net/view/2938321/3620189#tool-3622338 to listen to the audio.
Potential Paths
The growth models started to impact my results, choices, and musical aesthetics. Somehow, rapid, rhythmical sequences started to interest me and several approaches to repetitions, some quite extreme, began to take shape. Those sequences materialise through the activation of connected nodes, where the focus is given to the speed and dynamics of the transitions. The duration of each node is crucial, but also the contrast between successive nodes. The graphs can be activated sequentially across all nodes or operated interactively. Finally, sound analysis methods can influence network growth, adjusting structure to the properties of the sound being generated.
The core concept of this approach is to establish a chain reaction in the flow of events that is derived from both the material and its relations. This method introduces a ‘caused by’ attribute, that became essential to the resulting music. The growth process leaves discernible traces of activity, where each algorithmic alteration impacts slightly differently, depending on its timing and position within the network. These pivot points can themselves be viewed as musical elements, open for further development or for triggering additional processes. Every algorithmic choice marks a point in a graph of potential paths, each further iterated upon. In practice, the graph undergoes multiple executions with varying node settings and probabilities to evolve, refine, and solidify the final form of the music.
Observing these processes from a distance, they may appear organised and controlled, with actions that seem to comment on or direct an already established flow. The primary focus here is on how these methods generate events, and their capacity to shape compositional spaces as the network expands or shifts in dimension.