[Next] [Previous] [Up] [Top] [Contents]
6.1 Example strategies for agent model development
6.1.3 Generalizing models of the agent.
This would be tried when the agent has two models which are well confirmed according to some criterion. Generalization entails taking the intersection of the antecedents and the union of the consequents of the two models as the elements of the antecedents and consequents of the new model. The new model would apply in a wider range of conditions including at least all of the conditions in which both of the models from which it is generalized would apply. It would also forecast the values of a wider range of variables.
Generalizations will not be accepted if the resulting set of consequent clauses are either contradictory or cannot be instantiated from the resulting set of antecedent clauses. One implication of this condition is that the antecedents cannot be empty unless all of the values set by the generalized consequent are constants.
Modelling Learning as Modelling - 23 FEB 98
[Next] [Previous] [Up] [Top] [Contents]
Generated with CERN WebMaker