Adding Learning to Knowledge-Based Systems: Taking the "Artificial" Out of AI

Abstract

Both, knowledge-based systems (KBS) development and maintenance require time-consuming analysis of domain knowledge. Where example cases exist, KBS can be built, and later updated, by incorporating learning capabilities into their architecture. This applies to both supervised and unsupervised learning scenarios. In this paper, the important issues for learning systems-memory, feedback, pattern formulation, and pattern recognition-are described in terms of an instance vector set, a prototype vector set, and a mapping between those sets. While learning systems can possess robustness, recency, adaptability, and extensibility, they also require: careful attention to example case security, correct interpretation of feedback, modification for uncertainty calculations, and treatment of ambiguous output. Despite the difficulties associated with adding learning to KBS, it is essential for ridding them of artificiality.

Citation

Schmoldt, Daniel L. 1997. Adding Learning to Knowledge-Based Systems: Taking the "Artificial" Out of AI. AI Applications 11(3): 1-7.