Publication View

Logic-based Learning in Conflict Simulation Domains (2007)

Abstract
It is infeasible to specify complex Multi-Agent Systems in advance. Machine Learning techniques enable agents to learn from and adapt to the dynamic environment. So far, Reinforcement Learning methods have been extensively used in Multi-Agent Learning. However successful this approach has been, it relies on an assumption that establishes strict limits in the system's scalability: Agents are not provided with background knowledge. They learn from scratch. It is well-known, on the other hand, that domain-knowledge reduces considerably the complexity of planning problem. We propose to directly incorporate domain knowledge in the reasoning and learning processes of logic-based agents. To test this idea, we plan to apply Explanation-Based Learning and Inductive Logic Programming techniques in Conflict Simulation domains. 1

Publication details
Download http://citeseerx.ist.psu.edu/viewdoc/summary?doi=?doi=10.1.1.26.1556
Source http://www.cs.york.ac.uk/~ea/koldo.ps.gz
Contributors CiteSeerX
Repository CiteSeerX - Scientific Literature Digital Library and Search Engine (United States)
Type text
Language English
Relation 10.1.1.134.2462, 10.1.1.27.4632, 10.1.1.102.5130, 10.1.1.32.4337, 10.1.1.9.3725, 10.1.1.137.3032, 10.1.1.21.3961, 10.1.1.26.4453, 10.1.1.26.3484, 10.1.1.51.6243