Distributed Learning: An Agent-Based Approach to Data-Mining

this paper only presents preliminary, propositional results which do not reflect the more complex aspects associated with first order learning. Software agents (Levy, Sagiv & Srivastava, 1994; Oates, Prasad & Lesser, 1994) are one response to the problem of using the vast amounts of information stored on networked systems. There are many types of software agent (Wooldridge & Jennings, 1995); however, agents are typically thought of as being 'intelligent' programs which have some degree of autonomy. We intend to design an open, flexible data-mining agent. A group of these agents will be able to co-operate to discover knowledge from distributed sources. The issues involved have been covered to some extent by researchers concerned with multi-agent machine learning. A number of multi-agent learning systems have been built. These include MALE (Sian, 1991), ANIMALS (Edwards & Davies, 1993) and ILS (Silver, 1990). The first of these was a homogeneous, blackboard-based system, while the others used a distributed problemsolving approach. All three used propositional learning methods. A number of researchers have addressed the problem of reconciling different views of similar data; these include Gams (1989), Brazdil & Torgo (1990), and Svatek (1995). Our high-level model is as follows. One or more agents per network node are responsible for examining and analysing a local data source. In addition, an agent may query a knowledge source for existing knowledge (such as rules or predicate definitions). The agents communicate with each other during the discovery process. This allows the agents to integrate the new knowledge they produce into a globally coherent theory. In addition, a supervisory agent, responsible for co-ordinating the discovery agents may exist. A graphical interface..