The Automatic Training of Rule Bases that Use Numerical Uncertainty Representations
This work addresses knowledge acquisition challenges for developers of rule-based systems with uncertainty, though it appears incremental as it builds on existing optimization and truth maintenance techniques.
The paper tackles the difficulty of manually assigning rule weights in systems using numerical uncertainty representations by proposing automatic training via numerical optimization, reporting preliminary test results that show improved efficiency when combined with truth maintenance.
The use of numerical uncertainty representations allows better modeling of some aspects of human evidential reasoning. It also makes knowledge acquisition and system development, test, and modification more difficult. We propose that where possible, the assignment and/or refinement of rule weights should be performed automatically. We present one approach to performing this training - numerical optimization - and report on the results of some preliminary tests in training rule bases. We also show that truth maintenance can be used to make training more efficient and ask some epistemological questions raised by training rule weights.