Relational Evidence Theory and Interpreting Schematics

A new relational learning algorithm, the Consolidated Learning Algo-rithm based on Relational Evidence Theory (CLARET) is presented. Here, two different approaches to evidential learning are consolidated in how they apply to generalising within relational data structures. Attribute-based discrimination (decision trees) is integrated with part-based interpretation (graph matching) for evaluating and updating representations in spatial domains. This allows an interpretation stage to be incorporated into the generalisation process. These components of the system are demonstrated in an on-line system for the recognition of hand drawn, schematic dia-grams and spatial symbols. The approach uses an adaptive representational bias and search strategy during learning by efficiently grounding the learn-ing procedures in the relational spatial constraints of their application. It demonstrates how inductive rule generation can be constrained by domain knowledge