Our goal is to investigate how complex cognitive behaviour in artificial systems can emerge through interacting with an environment, and how, by becoming sensitive to the properties of the environment, such systems can autonomously develop effective representations. The underlying hypothesis is that perception is an active process; even in the absence of overt behaviour, perception involves prediction, and the need for making better predictions is what drives the development of useful representations and cognitive structures.
We will explore these issues within the realm of music cognition. Music is an ideal domain in which to investigate cognitive behaviour, since it is a universal phenomenon containing complex abstractions and temporally extended structures. As music is self-referential there are no externally determined semantics; the appropriate segmentation of the stream of sounds depends upon the structure of the signal itself, rather than the need to individuate objects in the external world.
By focusing on music cognition we can directly address problems such as the autonomous development of representations and processes that support the characterisation of events and event sequences, the development of categories and useful abstractions, the representation of situational context, interactions between long-term knowledge structures and working memory, the role of attention in optimising processing with respect to the current object of interest, the representation of temporal expectancies, and the integration of events across many different time scales.
We will investigate music cognition through perceptual experiments and computational modelling studies, embodying our understanding in the construction of an emergent interactive music system, which will learn to develop representations and expectations in response to the music it experiences, and will use these predictions to generate actions in the form of appropriately timed and pitched sounds.
Funding SchemeSTREP - Specific Targeted Research Project
1018 TV Amsterdam