Rapid insect-like visual learning algorithms
Investigators: Professor TS Collett and Professor P Husbands
Researchers: Dr. N. Hempel de Ibarra, Dr A Philippides, Mr L Smith
In marked contrast to current artificial systems, navigating insects learn complex visual sequences in remarkably few trials. An important part of this ability is the use of innate behaviours to structure and guide the learning process. Given their relatively small neural resources, insects have good reason to place a premium on finding such simplifying strategies. Enough is now known about these behavioural strategies to attempt a formal understanding of this type of learning. The central aim of this proposal is to provide such an understanding using detailed information theoretic analyses of data acquired during insect and robotic studies. This will inform the design of novel learning algorithms, to be demonstrated in vision-based navigation and other sequential learning application, as well as the development of more general principles for the design and understanding of adaptive systems.
Funding: EPSRC
Starts: 1 January 2005 Ends: 30 June 2008