Knowledge acquisition will always remains a key problem in the development of knowledge-based systems. With this motivation, a divergent number of methodologies and associated issues are appearing in the literature. This paper looks at how certain induction theories and methodologies conform to the requirements of knowledge acquisition from the light of practical experience. Experiments with perceptrons versus 'idiot' Bayes are reported, and an evaluation of Valiant's learning framework is made that yields improved bounds for reliable learning. Finally, some requirements for inductive knowledge acquisition methodologies are summarized.
- knowledge acquisition
- Valiant's framework