Neural network models for the study of post-tonal music

Abstract

Neural networks are used to study two issues pertaining to atonal music. In the first part of the paper, feed-forward neural networks, using a variant of the backpropagation learning algorithm, try to learn a variety of abstract theoretical constructs from pitch-class set theory. First, learning the properties of individual sets is studied. Then a network's ability to learn various relationships between sets is examined. Based on the behavior of the network during learning, conclusions are drawn with regard to perceptual issues relating to pcset theory. In the second part of the paper, an interactive activation and competition (IAC) network is used to parse a musical passage into analytical objects. The paper concludes with suggestions for further research.