J.Kleer -- "An assumption-based TMS", Artificial Intelligence (Holland), vol.28, n.2, pp.127-162, 1986  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of examples, is included in the rule R 0 1 . The method we propose, does not make any explicit distinction between parameter and structure identification, but both are obtained simultaneously in the learning algorithm. The identification model idea is based on the basic principles of an ATMS [De Kleer 86] We shall use the assumption and premise, contradiction and ATMS node terms. 2 ATMS Truth Maintenance Systems (TMS) are common tools in artificial intelligence for managing logical relationships between beliefs and statements. A more recently proposed ATMS [De Kleer 86] Zurita 94] maintains ....

....principles of an ATMS [De Kleer 86] We shall use the assumption and premise, contradiction and ATMS node terms. 2 ATMS Truth Maintenance Systems (TMS) are common tools in artificial intelligence for managing logical relationships between beliefs and statements. A more recently proposed ATMS [De Kleer 86] Zurita 94] maintains multiple sets of beliefs simultaneously, thus allowing inferences in multiple contexts at the same time. The problem solver builds records of all the inferences mades (justifications) and hypotheses it introduces (assumptions) The task of the ATMS is to efficiently ....

De Kleer, J. An assumption-based TMS. Artificial Intelligence, 28, 1986, 127-162.

....ii) W [OE T p , and iii) 8d 2 T, W [OE fT Gammadg 6 p. An argument for p is then a minimal consistent subset of defeasible rules augmented by our knowledge W that implies p. Notice that this notion of argument is very similar to the notion of environment used in the terminology of the ATMS (De Kleer, 1986). We will give later how to compute arguments using ATMS. Arguments are also used and extended to prioritized knowledge bases in (Benferhat et al. 1993b) Cayrol, 1995) in order to handle inconsistent knowledge bases. To formally define the notion of specificity, we first need to introduce the ....

....and S 0 Gammaarguments, namely how to compute arguments, ffl how to check if an argument is more specific than another, and ffl how to check if an argument counterargues another argument. For the first point we can use an ATMS tool. An ATMS (assumptionbased truth maintenance system) (DeKleer, 1986) is devoted to hypothetical reasoning. This system uses two kinds of propositional symbols, the assumption ones and the non assumption ones. An ATMS is able to determine 16 under which set of assumptions a given proposition p is true. This set of assumptions when it is minimal (with respect to ....

J. De Kleer (1986), An assumption-based TMS. Artificial Intelligence, 28, 127-162.

....micro world belief revision [Foo and Rao, 1991] etc. were vigorous for the last decade [G ardenfors, 1992a] Other non logic based dynamic nonmonotonic reasoning systems have also been proposed. 3 The most well known group of research is the truth maintenance systems [Doyle, 1979; de Kleer, 1986a] In the connectionists community, there have been recent studies in the direction of approximating symbolic reasoning using neural networks [Takagi and Hayashi, 1989; Kosko, 1990; Chan et al. 1993] In contrast to traditional logic based approaches, this group provides an alternative view for ....

....rules can be expressed as normal defaults as originally claimed by Reiter [Reiter, 1980] 1.1. 3 Truth Maintenance System Another class of systems capable of performing nonmonotonic reasoning when belief changes are carried out is Truth Maintenance Systems [Doyle, 1979; Thompson, 1979; de Kleer, 1986a; de Kleer, 1986b; Dixon and de Kleer, 1989] 9 A typical truth maintenance system is used to record and maintain the consistency and reasons for beliefs. It is designed to work with an external problem solver. The problem solver includes all domain knowledge and inference procedures. While the ....

[Article contains additional citation context not shown here]

J. de Kleer. An assumption based TMS. Artificial Intelligence J., 28:127--162, 1986.

....1989 ] Having briefly mentioned non monotonic logic, the remainder of this paper is dedicated exclusively to monotonic truth maintenance systems. There are several reasons for this. First, most of the development in truth maintenance algorithms, and de Kleer s ATMS algorithm in particular [ de Kleer, 1986a ] concern monotonic systems. Second, most practical applications of truth maintenance systems involve monotonic systems (e.g. qualitative simulation, fault diagnosis, and applications to search) Furthermore, monotonic truth maintenance systems provide a solid foundation upon which to build ....

....always included) and only a few premises are being changed one can arrange for the changing premises to migrate to the top of the stack so that only small retractions and additions are done when switching between contexts. ATMS like Implementations I will initially describe de Kleer s ATMS [ de Kleer, 1986a ] as an alternative implementation of the generic TMS interface and compare the ATMS implementation with the BCP implementation. For many applications the only difference between the ATMS and BCP implementations is there relative efficiency as implementations of the generic interface. However, ....

[Article contains additional citation context not shown here]

J. de Kleer. An assumption-based tms. Artificial Intelligence, 28:127--162, 1986.

.... Unfortunately, despite the long acknowledged utility of DDB, it fails to provide a coherent account of DDB, especially a lack of a coherent framework on the efforts on DDB and learning methods (e.g. Explanation Based Learning (EBL) Minton et al. 1989] or Justification Based Learning (JBL) [de Kleer 1986] ) Local algorithms use hill climbing techniques to find a solution quickly but are nonsystematic in that they search the whole space based on a probabilistic sense. These nonsystematic algorithms appear to be empirically effective because their ability to follow the local gradients based on ....

....in an order which is different from the one which has been used 6 to assign them. Both conflict directed backjumping and dynamic backtracking are built on the notion of eliminating explanation (defined by [Ginsberg 1993] or nogood justification (which comes itself from the TMS community [de Kleer 1986]) In this paper, the objects to be considered are called failure explanations because of their relations with explanation based method for failure driven learning. 3.1 Constraints and Failure Explanations: Preliminaries In order to understand the approach easily, we begin with some preliminary ....

de Kleer, J., An Assumption-Based TMS, Artificial Intelligence , 28, pp. 127-162, 1986

....degree of belief computation can be visualized diagrammatically. For example, suppose node o has eight inputs from o 1 ; o 8 . Three of them, o 1 , o 4 and o 7 , give positive inputs in descending order of degree of belief values (i.e. deg[o 1 ] deg[o 4 ] deg[o 7 ] another three nodes, o 6 , o 8 and o 2 , give negative inputs in descending order of degree of belief values, o 3 and o 5 have a proposition value of (0, 0) with deg[o 3 ] deg[o 5 ] and they give zero inputs; this is illustrated graphically in figure 1. If the sum of all positive inputs E is greater than the sum of all ....

.... g ( tg, fg) deg[g] 1, 0) normally Using the computation functions in definition 4, node value of node d is computed as follow: E = E a E b E c = 1 Theta 3 1 Theta 3 0 Theta 3 = 6 I = I a I b I c = 0 Theta 1 3 0 Theta 1 3 1 Theta 1 3 = 1 3 NET = E Gamma I = 6 Gamma 1 3 = 5 2 3 1 Therefore, the proposition value of node d, t d ; f d ) 1, 0) Since deg[a] deg[b] strongly weakly) and E a = 3 1, deg[d] deg[a] strongly. The node value of node g is computed using computation functions in definition 5. There are two rule links: one ....

[Article contains additional citation context not shown here]

J. de Kleer. An assumption based TMS. Artificial Intelligence J., 28:127--162, 1986.

.... prime covers are a way for either exhaustively, or concisely, representing the causes of failure of a system [13, 9] In automated reasoning, a prime cover can be either a set of minimal demonstrations, or the normalization of a formula into conjunctive normal form for further computations [12, 15, 16, 23, 18]. The problem of computing (irredundant, minimal) prime covers of Boolean functions was first addressed by Quine in the 1950 s [21] Since that date, efforts have been done to elaborate efficient prime cover computation procedures [22, 17, 2, 24, 25, 14, 3] For some problems, for instance for ....

J. De Kleer. An Assumption-based TMS. Artificial Intelligence, 28:127-162, 1986.

....on the constraint store return a con ict set , a set of labels that are involved in an inconsistency of the constraint store. The con ict set will be empty if the store is consistent. In the considered case, adaptation of constraint hierarchies relies on truth maintenance techniques (cf. [2, 3, 10]) that are used to adapt CHR derivations [18] More formally, adaptation works as follows: The initially given constraint store C is processed using rules for fd constraint solving. This results in a constraint store D. Then, the fd constraints justi ed by a set of labels M have to be eliminated ....

Johan de Kleer. An assumption-based TMS. Articial Intelligence, 28:127-162, 1986.

....of that theory. 3 We use the term minimal as per Definition 11. Definition 12 (Prime implicates) C is a prime implicate for Sigma iff Sigma j= C, and for no proper subset C 0 of C does Sigma j= C 0 . At the core of the well known assumption based truth maintenance system (ATMS) (de Kleer, 1986) is the computation of certain prime implicates of a propositional Horn theory, Sigma (Reiter and de Kleer, 1987) Thus, the ATMS contains a propositional consequence finding procedure for Sigma. In this discussion, we refer to the ATMS in the broadest context, to include its extensions beyond ....

....set of literals from which explanations are composed. It is equivalent to our set of observable and achievable literals when abduction is applied to test generation. It has long been known that there may be exponentially many abductive explanations for a given literal ( McAllester, 1985) (de Kleer, 1986)) and so listing them all would take exponential time. For test generation, we are often uninterested in listing all tests as explained by Property 1 above. Even if we were, by Property 3, we would be unlikely to have an exponential number of tests. Assuming, there are not an exponential number of ....

J. de Kleer (1986). An assumption-based TMS, in Artificial Intelligence Journal 28:127--162.

....has become a standard method for supplying semantics to nonmonotonic logic programs, there is a considerable interest in automating its computation. The earliest methods for nding stable models were based on the truth maintenance system of Doyle [15] and the assumption based TMS of deKleer [11], see [16, 17, 50] One of the rst algorithms that took advantage of the well founded semantics [65] was the 1 one presented in [29] and generalized in [30] Lately, more specialized algorithms have been developed [2, 63, 7, 12, 8, 18] However, they all su er from exponential space complexity ....

J. de Kleer. An assumption-based TMS. Articial Intelligence, 28:127{ 162, 1986.

....the same contradictions repeatedly. The general term for this behavior is thrashing [Mac92, SS77] There are several distinct kinds of thrashing. DeKleer identified four of them: futile backtracking, rediscovering contradictions, incorrect variable ordering, and rediscovering inferences [dK86] Of these four pathologies, only two can occur in DPL: 1. Futile backtracking occurs when DPL hits a dead end in the search space and backtracks to a choice that was not in any way responsible for the failure. There are two main approaches for avoiding this problem. The first approach, called ....

....the same contradictions repeatedly. The general term for this behavior is thrashing [Mac92, SS77] There are several distinct kinds of thrashing. DeKleer identified four of them: futile backtracking, rediscovering contradictions, incorrect variable ordering, and rediscovering inferences [dK86] Of these four pathologies, only two can occur in DPL: 1. Futile backtracking occurs when DPL hits a dead end in the search space and backtracks to a choice that was not in any way responsible for the failure. There are two main approaches for avoiding this problem. The first approach, called ....

[Article contains additional citation context not shown here]

Johan de Kleer. An assumption-based TMS. Artificial Intelligence, 28(2):127-- 162, March 1986.

....truth of a given proposition follows logically from a given set of premises and from the set of constraints, and to keep this information current. Two primary approaches to TMS implementation have been proposed: the JTMS (Justificationbased TMS) 13, 20, 22] and the ATMS (Assumption based TMS) [5]. Each design implements the main functionality in a different way. A JTMS starts with the given premise set and attempts to identify all provable propositions, hoping that a proof will be derived for the target proposition. An ATMS, on the other hand, maintains for each proposition a collection ....

....(The weights can also be used as a guide for additional measurement that should delineate between the different diagnoses. This example illustrates the efficiency of the belief revision process when the special structure of the problem is exploited. By contrast, handling this problem using ATMS [5] may exhibit exponential behavior. A similar algorithm exploiting the framework of probabilistic networks is given [16] 6 ATMS Labeling In this section we focus on the primary ATMS functionality, namely, finding one or all minimal instantiations of assumption variables in a given network of ....

J. de Kleer. An assumption-based TMS. Artificial Intelligence, 28(2):127--162, 1986.

....and more information is collected. Therefore our beliefs about the problem changes all the time and each of these changes is caused by some newly observed facts. Up to date, many approaches have been proposed to make such inference. Among them, the assumption based truth maintenance system (ATMS) [6] provides an attractive mechanism to maintain and update the belief set. The ATMS is a symbolic reasoning technique used in the artificial intelligence domain to deal with problems by providing dependent relations among statements during inference. This technique has been used in many areas such ....

....in nodes in extended incidence calculus. In section 5 we will briefly discuss how to provide justifications from extended incidence calculus. In the concluding section, we summarize the paper. 1.1. The basic reasoning model in the ATMS The truth maintenance system (TMS) 8] and later the ATMS [6] are both symbolic approaches to producing a set of statements in which we believe. The basic and central idea in such a system is that for each statement we believe in, a set of supporting statements (called labels or environments generally in the ATMS) is produced. A set of supporting statements ....

[Article contains additional citation context not shown here]

de Kleer,J., An assumption-based TMS. Artificial Intelligence 28 (1986) 127-162.

....significantly disagree in this case we may regard that predicting the weather for that day is subject to a high degree of uncertainty and so prone to error. 20 6. 2 Acting on inconsistency There are various proposals for resolving inconsistency including truth maintenance systems [Doy79, Kle86] knowledgebase merging [BKMS92, CH97, KP98, KP99, LS98, CH99] and belief revision (starting with [AGM85, Gar88] and with developments including iterated belief revision [DP97, Leh95] and relationships with database updating [KM92] For a review of belief revision theory see [DP98] Yet we may ....

J De Kleer. An assumption-based TMS. Artificial Intelligence, 28:127--162, 1986.

....2 (Figure 9) SAB required 1.27 seconds for the hard problem instance, while 0.22 seconds for the easy one. 6 Discussion and related work The common approach to model based diagnosis has widely been based on the language of propositional logic [ 7 ] and the reasoning machinery of the ATMS [ 5 ] . The main advantage of using constraint network techniques for diagnosis is that it allows the structure of the system to be effectively exploited and the complexity of computation to be bounded in advance. For example, the circuit in Figure 3 may have an exponential number of minimal conflict ....

J. de Kleer. An assumption-based tms. Artificial Intelligence, 28:127--162, 1986.

....a rule in P with variables x entail(x; CB ; H : H 0 G) CB A and CB A 6= false hH 0 CU ; CB i 7 hCU B u ; unify(C B A B b )i assuming that CB = unify(C B ) holds. Dependencies In principle, it is possible to manage and maintain dependencies in CHR derivations like ATMS labels [7, 15]. We suggest a simpler approach, where the labels of the derived constraints are the union of the labels of the constraints they depend on. Before we are able to define dependencies between constraints or computation steps with respect to a CHR derivation, we have to consider these CHR ....

Johan de Kleer. An assumption-based TMS. Artificial Intelligence, 28:127--162, 1986.

.... in (Wahlster and Kobsa, 1989) Furthermore, the beliefs of the user are rarely captured adequately by a static model: in order to handle the necessary dynamic nature, a truth maintenance system (either justification based (Doyle, 1979) as used in TRUMP (Bonarini, 1987) or assumption based (de Kleer, 1986) as used in GUMS (Finin, 1989) is required. Dynamic update of the user model during a dialogue involves a number of challenging tasks: key amongst these are belief ascription and plan ascription. Though clearly related, the former concentrates on inference drawn at the level of the individual ....

.... less, then more may be available for another (say, argument depth) This claim is consistent with results in cognitive psychology, demonstrating that tasks which place a load on centralised cognitive processing detract from the ability to perform further concurrent processing (Kahneman, 1973) Baddeley, 1986). It also follows in the spirit of Sperber and Wilson s theory of relevance which rests upon a key assumption that one fact is more relevant to the hearer than another if the hearer is called upon to perform less processing in order to derive that fact (Sperber and Wilson, 1986) in the current ....

, pp127-162

....In our approach ATMS techniques are used: Each initially given constraint is marked with an identifier presenting its justification. Each rule application and derived constraint is marked with a justifying set of constraint identifiers (see Figure 3) In contrast to classical ATMS approaches [6, 23] one justification and not all (minimal) justifications are maintained. 2.1 Requirements for Adaptations A closer look to the operational semantics of CHRs will show the requirements which are necessary for an adaptation of CHR applications. The operational semantics of CHRs is given by a ....

Johan de Kleer. An assumption-based TMS. Artificial Intelligence, 28:127--162, 1986.

.... in (Wahlster and Kobsa, 1989) Furthermore, the beliefs of the user are rarely captured adequately by a static model: in order to handle the necessary dynamic nature, a truth maintenance system (either justification based (Doyle, 1979) as used in TRUMP (Bonarini, 1987) or assumption based (de Kleer, 1986) as used in GUMS (Finin, 1989) is required. Dynamic update of the user model during a dialogue involves a number of challenging tasks: key amongst these are belief ascription and plan ascription. Though clearly related, the former concentrates on inference drawn at the level of the individual ....

.... less, then more may be available for another (say, argument depth) This claim is consistent with results in cognitive psychology, demonstrating that tasks which place a load on centralised cognitive processing detract from the ability to perform further concurrent processing (Kahneman, 1973) Baddeley, 1986). It also follows in the spirit of Sperber and Wilson s theory of relevance which rests upon a key assumption that one fact is more relevant to the hearer than another if the hearer is called upon to perform less processing in order to derive that fact (Sperber and Wilson, 1986) in the current ....

, pp127-162

No context found.

J.Kleer -- "An assumption-based TMS", Artificial Intelligence (Holland), vol.28, n.2, pp.127-162, 1986

No context found.

J.Kleer -- "An assumption-based TMS", Artificial Intelligence (Holland), vol.28, n.2, pp.127-162, 1986

No context found.

Kleer, J. de (1986). An Assumption-Based TMS. In: Artificial Intelligence 28, North-Holland, Amsterdam,1 986, pp.:0127-0162.

No context found.

J.Kleer -- "An assumption-based TMS", Artificial Intelligence (Holland), 1986, 28(2), 127-162

No context found.

Kleer, J. 1986. An assumption-based TMS. Artificial Intelligence (Holland). 28(2):127-162

No context found.

J. D. Kleer, "An assumption-based TMS," Artificial Intelligence, vol. 28, pp. 127--162, 1986.

First 50 documents  Next 50

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