In algorithmic music composition, a simple technique involves selecting notes sequentially according to a transition table that specifies the probability of the next note as a function of the previous context. I describe an extension of this transition table approach using a recurrent autopredictive connectionist network called CONCERT. CONCERT is trained on a set of pieces with the aim of extracting stylistic regularities. CONCERT can then be used to compose new pieces. A central ingredient of CONCERT is the incorporation of psychologically-grounded representations of pitch, duration, and harmonic structure. CONCERT was tested on sets of examples artificially generated according to simple rules and was shown to learn the underlying structure, even where other approaches failed. In larger experiments, CONCERT was trained on sets of J. S. Bach pieces and traditional European folk melodies and was then allowed to compose novel melodies. Although the compositions are occasionally pleasa...

Few types of signal streams are as ubiquitous as music. Here we consider the problem of extracting essential ingredients of music signals, such as well-defined global temporal structure in the form of nested periodicities (or meter). Can we construct an adaptive signal processing device that learns by example how to generate new instances of a given musical style? Because recurrent neural networks can in principle learn the temporal structure of a signal, they are good candidates for such a task. Unfortunately, music composed by standard recurrent neural networks (RNNs) often lacks global coherence. The reason for this failure seems to be that RNNs cannot keep track of temporally distant events that indicate global music structure. Long Short-Term Memory (LSTM) has succeeded in similar domains where other RNNs have failed, such as timing & counting and learning of context sensitive languages. In the current study we show that LSTM is also a good mechanism for learning to compose music. We present experimental results showing that LSTM successfully learns a form of blues music and is able to compose novel (and we believe pleasing) melodies in that style. Remarkably, once the network has found the relevant structure it does not drift from it: LSTM is able to play the blues with good timing and proper structure as long as one is willing to listen.

Communication signals in many animal species (including humans) show a surprising amount of variety both across time and at any one instant in a population. Traditional accounts and simulation models of the evolution of communication offer little explanation of this diversity. Sexual selection of signals used to attract mates, and the coevolving preferences used to judge those signals, can instead provide a convincing mechanism. Here we demonstrate that a wide variety of "songs" can evolve when male organisms sing their songs to females who judge each male's output and decide whether or not to mate with him based on their own coevolved aesthetics. Evolved variety and rate of innovation are greatest when females combine inherited song preferences with a desire to be surprised. If females choose mates from a small pool of candidates, diversity and rate of change are also increased. Such diversity of communication signals may have implications for the evolution of brai...

Abstract. In general music composed by recurrent neural networks (RNNs) suffers from a lack of global structure. Though networks can learn note-by-note transition probabilities and even reproduce phrases, they have been unable to learn an entire musical form and use that knowledge to guide composition. In this study, we describe model details and present experimental results showing that LSTM successfully learns a form of blues music and is able to compose novel (and some listeners believe pleasing) melodies in that style. Remarkably, once the network has found the relevant structure it does not drift from it: LSTM is able to play the blues with good timing and proper structure as long as one is willing to listen. A note to referees: As the output of the network is musical, it is difficult to gain an appreciation by just reading the paper. The text references a URL containing helpful musical examples (www.idsia.ch/˜doug/blues/index.html). If this paper is accepted at ICANN, the authors look forward to playing some of these passages as part of the presentation. This note will be deleted before publication! The paper is 6 pages long without this note. 1

This paper proposes a representation for univoiced musical sequences, and tests experimentally our hierarchical artificial neural model on a musical time series --- the third voice of the sixteenth four-part fugue in G minor of the Well-Tempered Clavier (vol. I) of J. S. Bach. The results obtained suggest that the model can perform efficiently on both recognition and discrimination of real musical sequences. It could recognize instances of a referential sequence --- the theme of the fugue --- in the presence of noise, and could also discriminate those instances out from the entire music. 1. Introduction In the domain of artificial neural models, recurrence is probably the most studied approach for analysing time series 1 . We can find in the literature several such models, varying from supervised models [26, 22, 11] to unsupervised ones [9, 14]. Our model is a recurrent one. It is based on the Kohonen's self-organizing map (SOM) [15]. It employs two SOMs endowed with time integrator...

In general music composed by recurrent neural networks (RNNs) suffers from a lack of global structure. Though networks can learn note-by-note transition probabilities and even reproduce phrases, attempts at learning an entire musical form and using that knowledge to guide composition have been unsuccessful. The reason for this failure seems to be that RNNs cannot keep track of temporally distant events that indicate global music structure. Long Short-Term Memory (LSTM) has succeeded in similar domains where other RNNs have failed, such as timing & counting and CSL learning. In the current study we show that LSTM is also a good mechanism for learning to compose music. We compare this approach to previous attempts, with particular focus on issues of data representation. We present experimental results showing that LSTM successfully learns a form of blues music and is able to compose novel (and we believe pleasing) melodies in that style. Remarkably, once the network has found the relevant structure it does not drift from it: LSTM is able to play the blues with good timing and proper structure as long as one is willing to listen.

We address the problem of musical variation (identification of different musical sequences as variations) and its implications for mental representations of music. According to reductionist theories, listeners judge the structural importance of musical events while forming mental representations. These judgments may result from the production of reduced memory representations that retain only the musical gist. In a study of improvised music performance, pianists produced variations on melodies. Analyses of the musical events retained across variations provided support for the reductionist account of structural importance. A neural network trained to produce reduced memory representations for the same melodies represented structurally important events more efficiently than others. Agreement among the musicians' improvisations, the network model, and music-theoretic predictions suggest that perceived constancy across musical variation is a natural result of a reductionist mechanism for p...

We describe a series of experiments using sequential neural networks to model the effect of contextual bias in music cognition. The model quantifies the strength and specificity of a virtual listener’s expectations while listening to functional tonal harmonic chord sequences. The network integrates pools of duple and triple metric units with pitch class representations of chords. The ’listener ’ is then exposed to new chord sequences. We interpret the output of each sequential vector as the expectation for the next event. By representing segregated duple and triple metric beat units we visualize the process of metric cognition, and the mutual reliance of metric and functional harmonic expectations in establishing a percept of meter and a context for expecting consequential harmonic activity. 1. Background Recent studies in cognition address the interdependence and mutual influence of multiple schemas in creating contexts. Schema based studies and models of music cognition [Leman, 1998] have applied Gestalt approaches to many aspects of the musical experience. However, few of these studies consider premonitory conditioning in schema selection. In this paper we show that a-priori contextual bias can strongly influence a sequential neural network model of music cognition and propose that this influence

In this paper, we argue that music cognition involves the use of acoustic and auditory codes to evoke a variety of conscious experiences. The variety of domains that are encompassed by music is so diverse that it is unclear whether a single domain of structure or experience is defining. Music is best understood as a form of communication in which formal codes (acoustic patterns and their auditory representations) are employed to elicit a variety of conscious experiences. After proposing our theoretical perspective we offer three prominent examples of conscious experiences elicited by the code of music: the recognition of structure itself, affect, and the experience of motion.

Many connectionist approaches to musical expectancy and music composition let the question of "What next?" overshadow the equally important question of "When next?". One cannot escape the latter question, one of temporal structure, when considering the perception of musical meter. We view the perception of metrical structure as a dynamic process where the temporal organization of external musical events synchronizes, or entrains, a listener's internal processing mechanisms. This article introduces a novel connectionist unit, based upon a mathematical model of entrainment, capable of phase- and frequency-locking to periodic components of incoming rhythmic patterns. Networks of these units can self-organize temporally structured responses to rhythmic patterns. The resulting network behavior embodies the perception of metrical structure. The article concludes with a discussion of the implications of our approach for theories of metrical structure and musical expectancy. Connection Science...