Designing a meta-learner by a neuro-fuzzy approach

Inductive learning mechanisms offer the tools for knowledge enlargement, but an analysis of common learning strategies reveals the limitations of base-learning methods. The objective of our research consists in defining a meta-learning framework which brings together a base-learner and a meta-learner with the aim of dynamically selecting a proper bias for the base-learner when a given task has to be tackled. Starting from the examination of meta-features (describing the properties of specific tasks) the meta-learner provides for a set of fuzzy rules. These represent an explicit form of meta-knowledge that can be employed to assess the most suitable bias in different learning domains. Contrasting with most of the meta-learning strategies exhibited in literature, our proposal adopts a single learning scheme - a neuro-fuzzy approach - for both the base-learner and the meta-learner.