MALEF: MultiAgent Learning Framework 1 MALEF: Framework for Distributed Machine Learning and Data Mining

Growing importance of distributed data mining techniques has recently attracted attention of researchers in multiagent domain. Several agent-based application have been already created to solve the data mining tasks. Most of these applications are based only on agentification of classic distributed data mining techniques. In this article we present a novel framework MALEF (MultiAgent Learning Framework) designed for both the agent-based distributed machine learning as well as data mining. Proposed framework is based on (i) the exchange of metalevel descriptions of individual learning process among agents and (ii) online reasoning about learning success and learning progress by learning agents. Abstract architecture enables agents to exchange models of their local learning processes. We introduce also a full range of methods to integrate these processes. This allows us to apply existing agent interaction mechanisms to distributed data mining tasks thus leveraging the powerful coordination methods available in agent-based computing. Furthermore it enables agents to implement the meta-reasoning capability to reason about their own learning decisions. We have evaluated this architecture in a simulation of real-world domains. This paper shows how to apply MALEF to distributed clustering problem and also illustrates how the conceptual framework can be used in practical system in which different learners may be using different datasets, hypotheses and learning algorithms. We describe our experimental results obtained using this system, review related work on the subject, and discuss potential future extensions to the framework. 2 Jan Tožička, Michael Rovatsos, Michal Pěchouček and Štěpán Urban