Ctr. for New Music and Audio Technol. CNMAT), Univ. of California at Berkeley, 1750 Arch St., Berkeley, CA 94709
Musical instruments, whether acoustic or electronic, are often nonlinear. So too are many compositional algorithms used to generate sequences of notes. Effective artistic use of such instruments and compositional procedures requires controllers that transform the musician's intention into the parameters that operate on the nonlinear system. Although a number of thorny theoretical issues remain unresolved, neural networks show considerable promise for the identification and control of nonlinear dynamical systems. Neural network or connectionist controllers have the further advantage of adapting to the particular or even idiosyncratic features of a individual musician's performance style. Connectionist controllers trained with back propagation learning that map performance gestures to the parameter space of sound synthesis and compositional algorithms are demonstrated. Many of the most interesting sounds and musical effects occur on the verge of chaos as opposed to the regions of full blown chaotic behavior. Operating on this border of chaos requires refined control and exposes a number of challenging research issues in nonlinear control theory.