Wirth, R. #1989#. Completing logic programs byinverse resolution. In: Proceedings of the Fourth European Working Session on Learning. Pitman. London. pp. 239#250.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....proposed by Muggleton[8, 10] uses the mechanism of inverting the resolution procedure in logic programming to construct clauses from given examples and in some cases, background knowledge. Many practical methods have been studied based on the inverse resolution(Muggleton[8, 10, 9] Wirth[13]) and these methods have received much attention as more practical approaches than those methods based on the criterion of identification in the limit by Gold[6] IR is expected to be a promising mechanism in inductive learning as is the resolution principle in logic programming. In recent years, ....

....inverse resolution into EFSs and discussed learning EFS language classes by using inverse resolution without background knowledge. As future work, larger language classes will be considered by using hierarchical learning concept or learning with background knowledge proposed in our previous papers[13, 16] in EFS framework. ....

R. Wirth: "Completing Logic Programs by Inverse Resolution", in Proceedings of the Fourth European Working Session on Learning, Pitman, Morgan Kaufmann, 1989, pp. 66--65.

....operators. It additionally includes what is called a saturation operator. This marks literals replaced by inverse resolution, and allows these to be used in further operations. This decreases it s sensitivity to the order of examples. ITOU [Rouveirol:92] is an extension of IRES, whilst LFP2 [Wirth:89] replaces CIGOL s three operators with two operators. Both operators use least general generalisation in conjunction with the absorption and intra construction operator. All the above systems still require an oracle, and cannot recover from overgeneralisations once made. GOLEM s algorithm is as ....

Wirth, R. "Completing Logic Programs by Inverse Resolution", In Proceedings of the Fourth European Working Session on Learning, Pitman, Montpellier, France, 1989, pp. 239-250.

....inverse of deductive rules of inference. Since the deductive rule of resolution is complete for deduction an inverse of resolution should be complete for induction. This idea of inverse resolution was first introduced for first order logic in [89] Several authors have expanded on these ideas [144, 49, 121, 136]. Four rules of inverse resolution were introduced in [79] Absorption: q A p A; B q A p q; B Identification: p A; B p A; q q B p A; q Intra construction: p A; B p A; C q B p A; q q C Inter construction: p A; B q A; C p r; B r A q r; C In these rules ....

....systems search typically either specific to general or general to specific. Incremental systems usually employ a mixture of these strategies as they may need to correct earlier induced hypotheses. Incremental ILP systems include MIS [125] CLINT [107] MOBAL [54] FORTE [118] RX [134] LFP [144], and CIGOL [89] Nonincremental systems include GOLEM [90] FOIL [102] FOCL [95] GRENDEL [24] CLAUDIEN [113] mFOIL [32] and LINUS [66] 10.1.2. Interactive Non interactive In interactive ILP, the learner is allowed to pose questions to an oracle (i.e. the user) about the intended ....

[Article contains additional citation context not shown here]

R. Wirth. Completing logic programs by inverse resolution. In EWSL-89, pages 239--250, London, 1989. Pitman.

....operators (absorption and identification) that are in principle capable of producing any Horn clause expression, but are restricted to unit clauses to reduce their complexity. Rouveirol and Puget [RP89] discuss the difficulties involved in removing this restriction for their IRES system. LPF2 [Wir89] also based on inverse resolution, is not restricted to unit clauses, but uses restrictions on the inverse substitutions that may be assumed. Restricted background knowledge In Horn clause logic, the generalization relationship between two clauses is easy to handle without background knowledge ....

Ruediger Wirth. Completing logic programs by inverse resolution. In Katharina Morik, editor, Proc. Fourth European Working Session on Learning (EWSL), pages 239 -- 250, London/San Mateo, CA, 1989. Pitman /Morgan Kaufmann.

....rarely complete in applications, this constraint is now generally believed to be over restrictive. ffl Strong bias of vocabulary. Present inductive systems construct hypotheses within the limits of a fixed vocabulary of propositional attributes. An increasing amount of Machine Learning research [24, 28, 1, 44, 35, 16, 19] is concerned with algorithms capable of inventing auxiliary predicates when insufficient background knowledge is provided. 2 Inductive Logic Programming A growing body of researchers have started to work on problems of inductive reasoning within the confines of pure Prolog. In this paper the ....

....derived, and two related operations were named the V operators (see Section 4.1) Another pair of operators ( W ) also based on inverting resolution, were shown to be capable of inventing predicates which were not available within the background knowledge (Section 5. 1) Banerji [1] Wirth [44] Ishizaka [16] Ling and Dawes [19] and Rouveirol and Puget [35] have also described related methods for inventing new predicates. However, although predicate invention has been demonstrated on large scale problems within a propositional setting [24, 25] this is not yet the case for any first ....

[Article contains additional citation context not shown here]

R. Wirth. Completing logic programs by inverse resolution. In EWSL-89, pages 239--250, London, 1989. Pitman.

....theory (CoLT) and the one on algorithmic learning theory (ALT) not to mention the workshops on particular topics of learning. European and national conferences accomplish the activities of the field. 2 Other learning problems that are formalized in logic are the inversion of resolution [Wirth, 1989] and the incremental inductive learning in the limit [Angluin and Smith, 1983] 1 Integrating machine learning into knowledge acquisition The first research project on machine learning in the Federal Republic of Germany, LERNER, was funded by the Ministry for Research and Technology (BMFT) ....

Wirth, R. (1989). Completing logic programs by inverse resolution. In Morik, K., editor, Proc. Fourth European Working Session on Learning (EWSL), pages 239 -- 250, London/San Mateo, CA. Pitman/Morgan Kaufmann.

....extremely important as the rules are used in a robot application to derive higher level concepts from sensor observations in real time. Our restructuring methods transform a program without changing the coverage of the original target concepts. They use inverse resolution (see, e.g. 14] 21] [27]) i.e. they implement the W operator (see [14] as inter construction for chain Datalog rules. Thus, our approach is closely related to the one proposed by Sommer [23] However, his method FENDER does not yield the result we need. During the restructuring process new predicates are invented. We ....

R. Wirth. Completing logic programs by inverse resolution. In K. Morik, editor, Proc. Fourth European Workong Session on Learning (EWSL), pages 239--250. Morgan Kaufmann, 1989.

....of these systems require intervention in the form of an oracle to approve and name new concepts. Unfortunately, we are not aware of any work that has systematically evaluated the competing approaches or the practical utility of predicate invention in general. Finally, it should be noted that Wirth (1989) has done some experiments using ILP techniques for the grammar induction problem. His system learns recognizers expressed as definite clause grammars. However, the presented results were rather preliminary, and no generalization experiments on larger scale corpora have been reported. Given the ....

Wirth, R. (1989). Completing logic programs by inverse resolution. In Proceedings of the European Working Session on Learning, pp. 239--250 Montpelier, France.

....be obtained by inverting deductive ones. Several approaches corresponding to different assumptions about the deductive rule and the format of background theory and examples have been proposed and investigated [7, 25, 36] The most interesting is based on the inversion of the resolution principle [29, 34, 38]. Although inverse resolution has been proposed in [22] as an inference system which consists of four rules (Absorption, Identification and two rules for the introduction of new predicate symbols) more specific forms of these have been implemented by having a two stage operation: first, inverse ....

R. Wirth. Completing Logic Programs by Inverse Resolution. In Proc. of 4th European Workshop Session on Learning, pages 239--250, 1989.

....restriction is due to efficiency reasons. 2. Conjunction of n Literals: For instance, RINCON s [Wogulis Langley 89] intermediate concepts consist of a deliberate number of literals. 3. Disjunctive Definition: Systems using the intraconstruction operator [Muggleton Buntine 88, Rouveirol 92, Wirth 89] construct predicates with disjunctive definitions. ffl Recursive Predicate: Several systems [Ling 91, Lapointe et al. 93, Ling 95] are designed to invent recursive new predicates. Ling states that recursive predicates are the only ones that are really necessary, because they cannot be ....

....predicates can express multi valued dependencies, i.e. dependencies that assign a set of values instead of only one value (as in the case of functional dependencies) to the values of the input variables. Reference Ext. Conj. 2 Lits. Disj. Rec. Deps. CIGOL [Muggleton Buntine 88] Theta LFP2 [Wirth 89] Theta ITOU [Rouveirol 92] Theta Banerji [Banerji 92] Theta RINCON [Wogulis Langley 89] Theta INPP [Ling 95] Theta CILP [Lapointe et al. 93] Theta CLINT CIA [De Raedt Bruynooghe 92] Theta FOCL [Silverstein Pazzani 93] Theta GOLEM [Muggleton 94] Theta CWS [Srinivasan et al. ....

[Article contains additional citation context not shown here]

Wirth R.: Completing Logic Programs by Inverse Resolution, in Proceedings of the Fourth European Working Session on Learning (EWSL-89), Pitman, London, pp.239-250, 1989.

....be a partial method of inverting Robinson s resolution rule of deductive inference. Together with Buntine, Muggleton implemented an algorithm for Inverse Resolution (IR) in a Prolog system named CIGOL [86] Subsequently, other implementations of IR have been reported by a number of workers (e.g. [116, 139]) It is useful to think of IR as a syntactic or proof based approach to the problem of generalisation within FOL, whereas RLGG is a semantic or model based approach. The CIGOL system was used experimentally in a demonstration which highlighted the need for relational learning [85] some of the ....

R. Wirth. Completing logic programs by inverse resolution. In EWSL-89, pages 239--250, London, 1989. Pitman. BIBLIOGRAPHY 198

....whereby entirely new predicates (neither present in E nor B) are induced. Predicate invention results in extending the vocabulary of the learner and may therefore facilitate the learning task. Although the idea of inventing new predicates is very appealing and some promising first approaches [30, 32, 46, 10, 44, 45] have been developed, the task of inventing new predicates is not well understood yet. ffl Single versus multiple predicate learning. In single predicate learning, the evidence contains examples for only one predicate and the aim is to induce a definition for this predicate; in multiple predicate ....

R. Wirth. Completing logic programs by inverse resolution. In K. Morik, editor, Proceedings of the 4th European Working Session on Learning. Pitman, 1989.

....(ILP) can hopefully provide advantages for NLP applications by increasing flexibility and limiting the amount of feature engineering required. Despite this fact, other than our own work, there has apparently been no application of ILP methods to language processing, with one early exception (Wirth, 1988, 1989). Over the last four years, we have explored the application of ILP to NLP. In particular, we have developed and extended the Chill system for acquiring natural language parsers (Zelle Mooney, 1993b, 1994b, 1996a, 1996b; Zelle, 1995) 1 This system has learned both syntactic and semantic ....

Wirth, R. (1989). Completing logic programs by inverse resolution. In Proceedings of the Fourth European Working Session on Learning, pp. 239--250. Pitman.

....lack of ambiguity in this sample task tends to compensate for the larger example space, resulting in excellent generalization to unseen cases. 4. 4 Related Work Most computational research in language acquisition has focused on the learning of syntax rather than search control or semantic parsing (Wirth, 1989; Berwick and Pilato, 1987; VanLehn and Ball, 1987; Berwick, 1985; Wolff, 1982; Liu and Soo, 1992) However, A number of language acquisition systems may be viewed as the learning of search control heuristics. Langley and Anderson (Langley, 1982; Anderson, 1983) have independently posited ....

Wirth, R. (1989). Completing logic programs by inverse resolution. In Proceedings of the European Working Session on Learning, pages 239--250. Montpelier, France: Pitman.

....and negative literals of C and D ; D Gamma be the same for D. Now if C D then C subsumes D and C Gamma subsumes D Gamma . 2 Sub unification The problem of inverting implication is discussed in a recent paper by Lapointe and Matwin [11] They note that inverse resolution [15, 14, 22, 24] is incapable of reversing SLD derivations in which the hypothesised clause is used more than once. In fact Plotkin [19] showed that the same problem appears in the use of relative least general generalisation of clauses. Lapointe and Matwin go on to describe sub unification, a process of matching ....

....Logic Programming and Machine Learning in general since all forms of generalisation involve inverting implication. The methods and algorithms described in this paper are derived from a first principles approach to the problem and extend previous approaches such as those using inverse resolution [15, 14, 22, 24] and relative least general generalisation [19, 2, 16] Although a first attempt has been made at this problem in a previous paper by Lapointe and Matwin [11] the author believes the approach taken in this paper to be more general and comprehensive. Various remaining problems with the square root ....

R. Wirth. Completing logic programs by inverse resolution. In EWSL-89, pages 239--250, London, 1989. Pitman.

....need a technique to invert resolution. 3. 2 Inverting One Resolution by Or introduction Other work on inverting resolution has primarily considered the problem of constructing one parent clause given the resolvent and the other parent clause (Muggleton Buntine, 1988; Rouveirol Puget, 1989; Wirth, 1989; Muggleton, 1990; Hume Sammut, 1991; Idestam Almquist, 1992; Rouveirol, 1992) Below we will describe how or introduction can be used to construct two parent clauses from only the resolvent. Let C and D be clauses, and the following clause R a resolvent of C and D: R = Cfl Gamma fAg) D ....

Wirth, R. (1989). Completing logic programs by inverse resolution. In Proceedings of the Fourth European Working Session on Learning. Pitman.

No context found.

Wirth, R. #1989#. Completing logic programs byinverse resolution. In: Proceedings of the Fourth European Working Session on Learning. Pitman. London. pp. 239#250.

No context found.

Wirth, R. (1989). Completing logic programs by inverse resolution. In Proceedings of the Fourth European Working Session on Learning (pp. 239--250). Montpellier, France: Pitman.

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