Inductive Hypothesis Validation and Bias Selection in Unsupervised Learning.

摘要: This paper approaches the importance of bias selection in context validating Knowledge Bases (KB) obtained by inductive learning systems. We propose a framework for automatic validation induced KBs based on capability shifting system. claim that this is useful not only when system has to validate its own results, but also human experts are available perform process. Experiments made using accuracy attribute prediction as performance goal. The unsupervised ISAAC [TC96] coupled with wrapper method [JKP94] search best bias. results support proposed ideas and suggest some future work seems interesting from both KB Machine Learning points view.