A.Ginsberg -- "A new approach to checking knowledge bases for inconsistency and redundancy", In Proceedings 3rd Annual Expert Systems in Government Conference, IEEE Computer Society, Washington, D.C., pp.102-111, October, 1987  Home/Search   Document Not in Database   Summary   Related Articles   Check

....hormone [6] Techniques exist for automatically generating test data sets. For example, the dependency network of a system can be used to determine inputs that will exercise all branches of the system. Sophisticated non monotonic techniques can be used to separate inputs into sensible subsets [5,18]. However, note that once an input suite is inferred, an expert still has to decide what are the appropriate outputs for those inputs. This may be a significant analysis task and, in practice, may only be practical for small systems. For all the above reasons, it is difficult and expensive to ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistency and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102-111, 1987.

....hormone [6] Techniques exist for automatically generating test data sets. For example, the dependency network of a system can be used to determine inputs that will exercise all branches of the system. Sophisticated non monotonic techniques can be used to separate inputs into sensible subsets [5,18]. However, note that once an input suite is inferred, an expert still has to decide what are the appropriate outputs for those inputs. This may be a significant analysis task and, in practice, may only be practical for small systems. For all the above reasons, it is difficult and expensive to ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistency and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....must be uncovered by a PSM reflecting over the generated proofs. In our example above (Figures 11 12) these were the ENV sets. Each such ENV set would be an interesting test case. Automatic test suite generation is systems which support inconsistencies have been implement by Ginsberg [46, 47] and Zlatereva [126, 127] The dependencies between 245 rules conclusions are computed and divided into mutually consistent subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....the criticism that this sample size is too small to make a general conclusion. However, we believe that we mutated that model over a sufficiently wide range to claim that our experimental results have some generality. EA is not the first validation algorithm defined using a multi worlds logic (see [9, 30] for others) However, to our knowledge, this is the first time that the limits to such an algorithm have been experimentally tested using a large test suite. We note that a lack of readily available models need not be a restriction to such testing. Any number of models and data sets can be ....

Ginsberg, A. A new approach to checking knowledge bases for inconsistency and redundancy in Proc. 3rd Annual Expert Systems in Government Conference. 1987. IEEE Computer Society.

....neuroendocrinological literature. Interesting, the detected faults were found using the data published to support those theories [16] An interesting variant on our validation approach are the automatic test suite 355 generation procedures offered by the dependency network approaches of Ginsberg [7, 8] and Zlatereva [27, 28] The dependencies between rules conclusions are computed and divided into mutually consistent subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, then the ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....unaware of other research into the complexity of temporal validation. However, for the non temporal case, our approach is similar in spirit to the nonmonotonic KBS validation research. For example, automatic test suite generation procedures based on dependency network are offered by of Ginsberg [Gin87] and Zlatereva [Zla93] The dependencies between rules conclusions are computed and divided into mutually consistent subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, then the ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....et al. 1987; Preece and Shinghal 1992) More sophisticated systems import the rule base into a truth maintenance system (TMS) and compute the worlds that include each conclusion. Zlatareva uses a JTMS variant (Doyle 1979; Zlatareva 1992; Zlatareva 1993) while Ginsberg uses an ATMS (DeKleer 1986; Ginsberg 1987; Ginsberg 1990) For example, if no ATMS context includes a conclusion, then it can never be made. With their detailed knowledge of inter dependencies, these TMS style validation tools can also critique a test suite or propose new test cases. For example, if the minimum labels of the worlds ....

Ginsberg, A. (1987). A new approach to checking knowledge bases for inconsistentcy and redundancy. Proc. 3rd Annual Expert Systems in Government Conference, IEEE Computer Society.

....so computational complex) Automatic tools such as HT4 can be a useful intelligent assistant for checking hypothetical ideas. An interesting variant on our external semantic testing approach are the automatic test suite generation procedures offered by the dependencynetwork approaches of Ginsberg [24, 25] and Zlatereva [71, 72] The dependencies between rules conclusions are computed and divided into mutually consistent subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, then the ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....must be uncovered by a PSM reflecting over the generated proofs. In our example above (Figures 12 13) these were the ENV sets. Each such ENV set would be an interesting test case. Automatic test suite generation is systems which support inconsistencies have been imple 390 ment by Ginsberg [48, 49] and Zlatereva [133, 134] The dependencies between page 17 of 56 if the patient received a ventricular urethral shunt then infection = e.coli ( 8) or klebsiella ( 75) Figure 14: The real heuristic knowledge within Figure 6. rules conclusions are computed and divided into mutually consistent ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and 1670 Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....such that mutually incompatible subsets of this dependency network are avoided. Such a representation can be used for validation. Thus dependency network can be used to determine inputs that will exercise all branches of the knowledge base. This is the basis of the validation systems by Ginsberg [20] and Zlatereva [58] However, note that once an input suite is inferred, an expert still has to decide what are the appropriate outputs for those inputs. In the case of poorly measured domains where there is no definitive oracle (e.g. neuroendocrinology) the correct outputs are unknown. Asking an ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....so computational complex) Automatic tools such as HT4 can be a useful intelligent assistant for checking hypothetical ideas. An interesting variant on our external semantic testing approach are the automatic test suite generation procedures offered by the dependencynetwork approaches of Ginsberg [29, 30] and Zlatereva [86, 87] The dependencies between rules conclusions are computed and divided into mutually consistent subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, then the ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....to testing) and earlier work on rule base analysis. Early approaches for rule base analysis carried out only verification or validation. A number of systems, such as the ONCOCIN Rule Checker Program (RCP) SSS82] CHECK [NPLP85, NPLP87] ESC (Expert System Checker) CS87] and KB Reducer [Gin87, Gin88] carry out only verification. Beyond their limitation to verification, these systems have additional weaknesses. RCP is limited to identification of static problems at the rule level, and cannot identify problems that result along longer reasoning chains. The CHECK system has better ....

....is a direct representation of the knowledge base and intermediate nodes represent intermediate hypotheses in a logical sense, but may have no particular meaning relative to the problem domain unless the system designer built that into the rules. 2. 2 KB Reducer As mentioned above, KB Reducer [Gin87, Gin88] uses the implied network of rules to do the rule base analysis. The process of knowledge base reduction involves calculation of all possible logically independent and minimal sets of inputs under which the knowledge base will conclude each assertion (each class) In order for the ....

[Article contains additional citation context not shown here]

A Ginsberg. A new approach to checking knowledge bases for inconsistency and redundancy. In Proceedings of Third Annual Expert Systems in Government Conference, pages 102--111, Washington, D.C., 1987.

....exclusive forks must be 640 managed in separate worlds. Further, the source code itself may have to be divided into portions (a.k.a. worlds) Each portion must be internally consistent. Automatic test suite generation is systems which support inconsistencies have been implement by Ginsberg [50,51] and Zlatereva [107,108] The dependencies between rules conclusions are computed and divided into mutually consistent 645 subsets. The root dependencies of these subsets represent the space of all reasonable tests. If these root dependencies are not represented as inputs within a test suite, ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

....such that mutually incompatible subsets of this dependency network are avoided. Such a representation can be used for validation. Thus dependency network can be used to determine inputs that will exercise all branches of the knowledge base. This is the core of the validation systems by Ginsberg [9] and Zlatereva [37] However, note that once an input suite is inferred, an expert still has to decide what are the appropriate outputs for those inputs. In the case of vague models (where there is no definitive oracle) the correct outputs are unknown. The remove model comprehension problem is a ....

A. Ginsberg. A new Approach to Checking Knowledge Bases for Inconsistentcy and Redundancy. In Proc. 3rd Annual Expert Systems in Government Conference, pages 102--111, 1987.

No context found.

A.Ginsberg -- "A new approach to checking knowledge bases for inconsistency and redundancy", In Proceedings 3rd Annual Expert Systems in Government Conference, IEEE Computer Society, Washington, D.C., pp.102-111, October, 1987

No context found.

A.Ginsberg -- "A new approach to checking knowledge bases for inconsistency and redundancy", In Proceedings 3rd Annual Expert Systems in Government Conference, IEEE Computer Society, Washington, D.C., pp.102-111, October, 1987

No context found.

A.Ginsberg -- "A new approach to checking knowledge bases for inconsistency and redundancy", In Proceedings 3rd Annual Expert Systems in Government Conference, IEEE Computer Society, Washington, D.C., pp.102-111, October, 1987

No context found.

Ginsberg, A. 1987. A new approach to checking knowledge bases for inconsistency and redundancy. In Proceedings of the 3rd Annual Expert Systems in Government Conference. 102-111. Washington, D.C., IEEE Computer Society.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC