Peter Tino

Peter Tino is Lecturer and part of the Natural Computation Group, School of Computer Science, University of Birmingham

We propose to endow approaches to evolutionary art with machine learning capabilities. Learning mechanisms will be employed to customise the system to a particular artist (group of artists) by enabling the machine learning component to monitor user activities and positive/negative preferences in the course of using the system. Different kinds of learning will be considered, e.g. supervised, unsupervised and reinforcement learning, depending on the nature and context of representations a particular learning subsystem operates on.

The learning component will be able to

• assist in creating new members of population,
• automatically categorise creative patterns of individual artists and relate such patterns across several artists/groups of artists,
• describe and monitor evolution of user tastes and preferences, etc.

---