Computational learning of observable moves can be likened to a human observer unfamiliar with the game of chess - by watching a game and the sequences of moves taken in the game, the observer can learn a lot about the rules. Watching a bishop for long enough for example would demonstrate that only diagonal moves are possible for this piece.
Dr Gregory will study how computers learn in this way and will develop new methods of numerical automated modelling. Many industrial and commercial applications use automated modelling to solve common problems such as automated planning, general game playing, interactive narrative production or workplace rostering.
Completion of the project will allow many different problems to be learnt simply from observation. Examples may include models to determine capacity limits for certain resources, or constraints for positioning items. Modelling may also be created to enable certain actions within a social network based on levels of observed friendship.