On the Development of Inductive Learning Algorithms: Generating Flexible and Adaptable Concept Representations

178 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1998.HCL achieves two functionalities: (1) flexibility in the representation, by increasing the complexity of the hypothesis comprised at each hierarchical layer, and (2) adaptability in the search for different representations, to know when to stop adding more layers in top of the hierarchical structure. Adaptability allows HCL to adjust the complexity of the representation by building few hierarchical levels when the concept is simple, and by increasing the number of levels as the difficulty of the concept grows higher. HCL is assessed experimentally using both artificial and real-world domains. Results show how HCL outperforms other models significantly when many intermediate concepts lie between the primitive features and the target concept. (Abstract shortened by UMI.).U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD