A probabilistic atms

From Semantic Portal Wiki

{{#vardefine:category|Publication}}{{#vardefine:templatename|i.publication}}{{#vardefine:package|smwbp_instance_templates}}

Edit

Reference: {{#vardefine:pagename|a probabilistic atms }}

1. [[]]

bibtex

{{#vardefine:pagename|A probabilistic atms }}{{#vardefine:key| }}

abstract: Truth maintenance systems (TMS) provide a method of improving theefficiency of search during problem solving. The problem solver usesthe TMS to record the reasons that facts are derivable so that factsneed not be rederived during the course of the search. De Kleer'sAssumption Based Truth Maintenance system (ATMS) overcomes thelimitations of many earlier systems, such as not being able to switchstates swiftly and not being able to consider multiple solutions to aproblem at once. We describe a probabilistic extension to the ATMS --An ATMS structure is augmented with a probability distribution overthe set of assumptions. A probabilistic model is then constructed inthe form of a Bayesian network from the ATMS structure. Theprobabilistic ATMS provides significant new functionality such as thederivation of the probability of a fact being derivable, the posteriorprobability over the assumptions given that a fact is derivable andthe most probable context in which a fact is derivable. Our techniquedoes not require the probability distribution of an assumption to beindependent of the distributions of other assumptions. As an exampleof the use of the probabilistic ATMS, we show that it can be appliedto construct probabilistic models to do multiple fault diagnosis. Thisgeneralizes some aspects of de Kleer and Williams' work on model baseddiagnosis. The probabilistic ATMS has been implemented in IDEAL, aBayesian network solver.

download:

• paper:
• slides: