Abstract
Past research in cartographic generalization has shown that algorithmic methods are well suited to handle narrow tasks, but appear to have limited potential so solve the entire generalization process comprehensively. Attempts to use systems based on explicit knowledge representation (e.g., rule-based or expert systems) also had relatively little success. The major limiting factor to explicit knowledge systems in generalization is the scarcity of formalized knowledge available. That is, knowledge acquisition (KA) forms the major bottleneck to progress of knowledge-based techniques.In this paper, we discuss what options are available for cartographic KA, assess their potential, and propose alternatives which are based on the integration of techniques of computational intelligence (CI) with interactive environments. CI methods have the advantage of avoiding explicit knowledge formulation. Interactive systems allow to keep the human expert in the loop and thus augment his/her productivity as well as the potential for KA. Two examples are presented for this approach. The first example focuses on knowledge acquisition by process tracing in interactive systems. A comprehensive KA methodology supported by inductive learning algorithms is proposed and its technical feasibility assessed by an experiment. In the second example, genetic algorithms are used as an optimization method for interactive control of parameter settings for line generalization operators in an approach termed ‘generalization by example”.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have