Realtime Stochastic Decision Making for Music Composition and Improvisation

Christopher Dobrian (University of California, Irvine, USA)

The theoretical and practical applications of the mathematics of probability in music composition, and the statistical bases of probabilistic compositional decision making, were proposed and described by Iannis Xenakis in his Formalized Music. [Xenakis 1963, 1971]  The development of electronic computation made many of those ideas realizable with great detail, complexity, and specificity in large-scale compositions.  It also led to various branches and elaborations of Xenakis’s ideas: statistical control of granular synthesis, visual/spatial representations of music parameter space, etc.  In recent years, faster computation has made enactment of these processes possible in real time, permitting probabilistic decision-making techniques to be employed in spontaneous improvisational contexts, and allowing real-time interactive navigation of multi-dimensional music parameter space.

Xenakis noted that musical textures, described by statistical properties, can change in a continuous or discontinuous manner, producing music that changes either by subtle nuance or by vigorous assertion. Vectoral computations of linear and exponential interpolations between different sets of probabilities can produce continuous changes, while a sudden shift from one set of probabilities to another can produce discontinuous change.  The balance of continuity and discontinuity, and the balance of entropic and negentropic conditions in music, can be guided by the musical taste and intuition of a composer or improviser, or indeed can be guided by a higher level computational system – an artificial intelligence or algorithm – to compose/improvise a larger musical structure over time.

This presentation will demonstrate programming techniques and composition/improvisation methodologies particularly well suited to the real-time implementation of probabilistic (i.e., stochastic) decision making.  These include exponential transitions from one set of probability vectors to another, and structures that permit an improviser to navigate interactively through sonic and musical virtual parameter spaces of stochastic textures.  Musical examples will be presented from the author’s compositions that employ stochastic compositional decision-making techniques, including Entropy for computer-controlled piano, Unnatural Selection for piano and synthesizer, and other works.

Christopher Dobrian is a Professor of Music at the University of Caliornia, Irvine.  He is the director of the Gassmann Electronic Music Studio and the Realtime Experimental Audio Laboratory (REALab), and is producer/director of the Gassmann Electronic Music Series.  He holds a Ph.D. in Composition from the University of California, San Diego, where he studied composition with Joji Yuasa, Robert Erickson, Morton Feldman, and Bernard Rands, and computer music with F. Richard Moore and George Lewis.  He is vice president of the Electronic Music Foundation, and is the author of the original technical documentation and tutorials for the Max and MSP programming environments by Cycling ’74.  His work in computer music focuses on the development of “artificially intelligent” interactive systems for composition, improvisation, and cognition.