Heuristic for multi-variable music state recommendations

Abstract

One current challenge in the CAIRA project, which is devoted to the relationship between creativity, logic and spontaneity in musical environments, is to provide logically provable recommendations to a creative agent based upon the input of two players in an ensemble, as well as the agent itself. We report on the meeting of this challenge herein. At present, the recommendations are proven valid based upon one variable for each player, and each possible combination of player variable ranks may be quickly computed and proven. When the agent enters a state that has not yet been proven, it uses a nearest-neighbor approach to determine which state to enter. However, as more player variables are added, recommendation error and lookup time increase. The focus of this thesis is to provide a method for providing more accurate guess states to the agent in a timely manner.