USING COMPILED KNOWLEDGE TO GUIDE AND FOCUS ABDUCTIVE DIAGNOSIS

Several artificial intelligence architectures acid systems based on '' deep'' models of a domain have been proposed, iii particular for the d iagnostic task. These systems have several advantages over traditional knowledge based systems, but they have a main limitation in their com putational complexity. One of the ways to face this problem is to rely on a knowledge compilation phase, which produces knowledge that can b e used more effectively with respect to the original one. In this pape r we show how a specific knowledge compilation approach can focus reas oning in abductive diagnosis, and, in particular, can improve the perf ormances of AID, an abductive diagnosis system. The approach aims at f ocusing the overall diagnostic cycle in two interdependent ways: avoid ing the generation of candidate solutions to be discarded a posteriori and integrating the generation of candidate solutions with discrimina tion among different candidates. Knowledge compilation is used offline to produce operational (i.e., easily evaluated) conditions that embed the abductive reasoning strategy and are used in addition to the orig inal model, with the goal of ruling out parts of the search space or f ocusing on parts of it. The conditions are useful to solve most cases using less time for computing the same solutions, yet preserving all t he power of the model-based system for dealing with multiple faults an d explaining the solutions. Experimental results showing the advantage s of the approach are presented.