The Dynamics of Learning and the Emergence of Distributed Adaption

This project developed fundamental theory and novel algorithms for adaptive learning in autonomous collective-agent systems. The first goal was to develop a new mathematical framework for analyzing the dynamical mechanisms that support learning in a range of novel information processing substrates. The second goal was to adapt this single-agent learning theory and associated learning algorithms to the distributed setting in which a population of autonomous agents communicate to achieve a desired group task. The results provide a mathematically sound basis for quantitatively measuring 1 the degree of individual-agent intelligence 2 the significance of agent-environment interaction and 3 the emergence of cooperation in agent collectives. A novel feature was that the theoretical approach synthesized recent results in the areas of machine learning, statistical inference, statistical mechanics, nonlinear dynamics, and pattern formation theory and then applied them to single-agent learning and adaptive learning in agent collectives. The central impact of the project will be its systematic and quantitative approach to predicting behavioral complexity in agents and in their environments. This will provide a foundation for measuring agent and agent-collective intelligence which, in turn, should allow for systematic engineering and monitoring of these systems.