Hybrid logic nets contain nodes that exhibit recurrent characteristics, in that a node output shows temporal tendencies even when the inputs are constant (Al-Dabass et al. 1999a and b). Given the behaviour trajectory of such a node it is required to abduct the values of the causal parameters. A new causal parameter estimation algorithm is given for abducting knowledge embedded within recurrent logic networks. The results show the new algorithm to be superior to previous algorithms when two of the causal parameters possess temporal behaviour.
David Al-Dabass, David J. Evans, Siva Sivayoganath