McIlraith, S., and Reiter, R. 1992. On tests for hypothetical reasoning. In W. Hamscher, L. C., and de Kleer, J., eds., Readings in model-based diagnosis. Morgan Kaufmann. 89--96.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to define the set of tests relevant for a given disorder d. This set contains manifestations not observed yet that are among predicted effects of d (whether predicted as present or absent) T(d) M(d) M(d) M M ) This definition is inspired from McIlraith and Reiter [25] and can be adapted to the possibilistic approach as follows: 22 T(d) m) min(max( M(d) m) M(d) m) 1 M (m) 1 M (m) 13) A test is relevant with respect to a set D of potential disorders if all disorders in D can be confirmed or discarded by testing m, ....

McIlraith S., Reiter R., "On tests for hypothetical reasoning", in: Readings in ModelBased Diagnosis (W. Hamscher et al., eds.), Morgan & Kaufmann, San Francisco, CA, 1992, 89-96.

....often called sequential diagnosis. The second feature refers to available knowledge, and says that all potential final diagnoses can be generated on the basis of the initial observations. This feature is exploited in the current set of methods here, applying the strategy of differential diagnosis [McIlraith Reiter, 1992]. 3 Structure and Strategy The main modelling components in the set are the expansion methods. They decompose a function into a set of lower level functions a function structure. Each function structure should have at least one control method. Each expansion method is guided by task features. ....

....ffl the prune explanations function rejects the subset of the explanations that cannot explain the additional observation. This method employs a strategy which McIlraith and Reiter have called differential diagnosis 4 . They express the intuitive idea behind the strategy in the following way ([McIlraith Reiter, 1992]) Given a set of potential diagnoses, a sequence of tests may be performed to iteratively reject diagnoses without the need for subsequent diagnosis generation steps. Following each test, the resulting set of hypotheses contains all and only the hypotheses to be entertained in further ....

[Article contains additional citation context not shown here]

Sheila McIlraith and Raymond Reiter. On tests for hypothetical reasoning. In W. C. Hamscher, J. de Kleer, and L. Console, editors, Readings in Model-based Diagnosis, pages 89--96. Morgan Kauffman, August 1992.

....in networks. More speculative is the use of our results to model the iterative behaviour of various existing diagnostic systems. Such systems iterate over multiple models (AbuHanna 1994) different abstraction levels (Mozetic 1991, Console and Torasso 1992) or request additional observations (Mcilraith and Reiter 1992). Our claim is that such iterative behaviour can be formalised in a uniform way through our results. This would yield insights in the differences and commonalities between such systems, and would make our anytime algorithms available to these existing systems. Further work is required to make ....

S. McIlraith and R. Reiter. On tests for hypothetical reasoning. In W. Hamscher, L. Console, and J. de Kleer, editors, Readings in Model-Based Diagnosis, pages 89--96. Morgan Kaufman, 1992.

....from positive and negative information. Future work lies in comparing our work to approaches from knowledge acquisition (e.g. human expert models, cooperative knowledge base tuning) machine learning (e.g. case base reasoning) as well as model based diagnosis (e.g. McIlraith and Reiter [13] study the design of tests whose outcomes confirm or refute a hypothesis) Huard and Freuder [11] test their knowledge elicitation method on random problems. However, they start from an idealized CSP P that must be approximated; our method is useful to find an unknown P , therefore random PCSPs do ....

S. McIlraith and R. Reiter. On tests for hypothetical reasoning. In W. Hamscher, L. Console, and J. de Kleer, editors, Readings in Model-Based Diagnosis, pages 89--96. Morgan Kaufmann, 1992.

....to restricting the conditional expression has yet been explored. 1.4.4 Diagnosis The goal of a search plan is the acquisition of information about the environment rather than some modification of the environment. This is a feature that search plans share with plans for diagnosis [RWC92, SW92, MR92] That actions may both be informative and effect changes in the world has been recognized as early as McCarthy and Hayes [MH69] and used in recent work on medical decision support [RWC92] That one may have to take multiple knowledge acquisition actions before one s knowledge is sufficient to ....

.... has been recognized as early as McCarthy and Hayes [MH69] and used in recent work on medical decision support [RWC92] That one may have to take multiple knowledge acquisition actions before one s knowledge is sufficient to take decisive action has been recognized in work by McIlraith and Reiter [MR92] on incremental diagnosis. Plans for incremental diagnostic tests such as proposed by McIlraith and Reiter are similar to search plans, except that they specifically exclude the possibility that the test actions will change the state of the world. As a consequence of expressing expected test ....

Sheila McIlraith and Raymond Reiter. On tests for hypothetical reasoning. Technical report, University of Toronto, 1992.

....standard first order formalism; and both [8] and [6] express only preferences among fault modes (roughly corresponding to our set INIT(N ) whereas we exploit the structure of the entire theory to express preferences. Our preferences are a way to select among competing solutions. Other work (e.g. [10]) tries to obtain new discriminating observations to select among competing solutions. Our preferences are a selection method that should be applied when it is no longer possible or desirable to obtain further discriminating observations. Future work: An obvious extension that is required to make ....

S. McIlraith and R. Reiter, `On tests for hypothetical reasoning', in Readings in Model-Based Diagnosis, eds., W. Hamscher, L. Console, and J. de Kleer, 89--96, Morgan Kaufman, (1992).

....for the specific problem of IC circuit testing, but there is little work on testing for rich dynamical systems such as the ones we examine here. The notion of a static test was briefly discussed in (Moore 1985, litmus example) and further developed for static systems in (McIlraith 1994; McIlraith Reiter 1992). We build directly upon the work in (McIlraith 1994) with the objective of developing a formal theory of testing for dynamical systems. Informally, a simple test comprises a set of initial conditions that may be established by the agent, together with the specification of a primitive sensing ....

McIlraith, S., and Reiter, R. 1992. On tests for hypothetical reasoning. In W. Hamscher, L. C., and de Kleer, J., eds., Readings in model-based diagnosis. Morgan Kaufmann. 89--96.

....of component parts. These domains did not require explicit representation of actions for the tasks they were trying to address. In contrast, there are many applications for which testing and repair require a sequence of actions which actually change the state of the world in important ways [8]. For example, the achievement of a test, such as biopsying a tumor or testing the spark plugs in a car involve actions which change the state of the world. Similarly, repair procedures such as those required in medical treatment or machinery repair change the state of the world and affect ....

S. McIlraith and R. Reiter (1992). On tests for hypothetical reasoning. In W. Hamscher et al. (ed.), Readings in model-based diagnosis, 89--96. Morgan Kaufmann.

....formal analysis of the problem of test generation in the AI literature. Our objective is to move beyond the specific problem of testing digital circuits and to examine the general problem of test generation for hypothetical reasoning, including diagnosis. In an earlier paper, McIlraith and Reiter (McIlraith and Reiter, 1992) provided a logical characterization of testing for hypothetical reasoning. They characterized tests in terms of the prime implicates P I ( Sigma) of Sigma. Since the ATMS computes (many) P I ( Sigma) in generating diagnoses, it was shown that some propositional tests could simply be read off ....

....necessary. Many tests are thus generated for free. While a nice result for ATMS based problem solvers, it is of limited use for hypothetical reasoning problem solvers that do not compute the prime implicates of Sigma. In this paper we take the logical characterization of tests introduced in (McIlraith and Reiter, 1992) and use it as a basis for examining the task of test generation. We augment and extend the framework to a first order characterization. Then we recast test generation as abduction. In so doing, we are able to apply the abundant research on abduction to gain insight into the generation of tests. ....

[Article contains additional citation context not shown here]

S. McIlraith and R. Reiter (1992). On tests for hypothetical reasoning, in W. Hamscher, L. Console and J. de Kleer, editors, Readings in model-based diagnosis, 89--96. Morgan Kaufmann Publishers, San Mateo, CA.

....circuits or related electro mechanical systems, where testing is nonintrusive and repair actions involve the simple replacement of component parts. There are many applications for which testing and repair require a series of actions which actually change the state of the world in important ways (McIlraith and Reiter, 1992). For example, the achievement of a test, such as biopsying a tumor or testing the spark plugs in a car require actions which change the state of the world. Similarly, repair procedures such as those required in medical treatment or machinery repair change the state of the world and affect ....

....has occurred. The potential actions can then be generated easily from the successor state axioms. 4 TESTING Given a theory of system behavior and a set of candidate diagnoses, we may wish to perform tests in order to discriminate these candidate diagnoses in some fashion. In previous papers ((McIlraith and Reiter, 1992), McIlraith, 1994) we have proposed a theory of testing for hypothetical reasoning which relates directly to the diagnosis literature. A test specifies some initial condition A which the tester establishes, and an observable O whose truth or instantiated value the tester is to determine from ....

[Article contains additional citation context not shown here]

S. McIlraith and R. Reiter (1992). On tests for hypothetical reasoning. In W. Hamscher, L. Console and J. de Kleer (ed.), Readings in model-based diagnosis, 89--96. Morgan Kaufmann Publishers.

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