Evidential reasoning neural networks
S.M. Mohiddin, Tharam Singh Dillon · 2002
This paper proposes an neural network architecture for evidential reasoning. This has been achieved by combining an extended multilayered neural network for learning rules and decision trees with a new interpretation. The new interpretation of the decision tree converts a decision tree into an evidential reasoning construct called hierarchy tree (HT). Fuzzy knowledge representation methods have been used in the HTs for approximate reasoning. Fusing the HT into a neural network it is shown that imprecision and ignorance can be handled.>