Abstract not found

Whilst the development of Learning Classifier Systems 1 has produced excellent results in some fields of application, it has been widely noted that problems emerge when seeking to establish higher levels of knowledge (see Barry (1993) for a relevant review). Tsotsos (1995) suggests that research into the operation of the Visual Cortex shows a hierarchical decomposition of processing more structured than a simple Subsumption Architecture arrangement. Whilst the LCS can provide both memory and planning by the use of Tags and Rule Chains, it provides a flat rule space. Various approaches have been taken to introducing structure to the LCS. We examine these approaches and identify three major lines of research: Multiple Interacting LCS, Single LCS with a structured population; and Structured Encoding of Rules. We illustrate that the first two of these areas have been interpreted quite differently, and seek to draw out common findings from the different approaches. We round off our examination of the area by a more detailed look at the work of Dorigo and Schnepf (1992), using a Hybrid Classifier System to examine the performance claims of Dorigo and Schnepf's architecture.

Abstract not found

We introduce a new approach to GA (Genetic Algorithms) based problem solving. Earlier GAs did not contain local search (i.e. hill climbing) mechanisms, which led to optimization difficulties, especially in higher dimensions. To overcome such difficulties, we introduce a "bug-based " search strategy, and implement a system called BUGS2. The ideas behind this new approach are derived from biologically realistic bug behaviors. These ideas were confirmed empirically by applying them to some optimization and computer vision problems. 1