Learning from observational data with prior knowledge

ion Domain layer Inference layer Task layer Strategy layer Figure 2: Degree of abstraction theoretical view When comparing the layers according to degree of abstraction, the above ordering suggests the idea of Fig. 2. However, the design of the strategy layer is, in practice, tightly bound with the problem domain, and can be viewed as more specific than the inference and task layers. Fig. 3 is thus more realistic. The inference and task layers have, on the other hand, many generic features which do not differ very much across problem domains; with a certain simplification, we can assume that they do not model the current, specific, task, but a generally defined, generic one. The reusability of inference and task layers, in the form of so-called generic task models, can save a notable amount of work compared to developing KBSs from scratch. In this respect, several versions of Generic Task Model (GTM) Library have been put into exploitation in the KBS development circles [TanHay93]. The set of GTMs contained in the library has been organized into a hierarchical structure, the top nodes of which being the notions of system analysis, Some of the GTMs are derived from generic parts of actual KBSs, others have been contrived “artificially”, by means of thorough analysis of human problem solving.