S. H. Muggleton and C. Feng. E#cient induction of logic programs. In Proceedings of the First Conference on Algorithmic Learning Theory, pages 368--381, Tokyo, 1990. Ohmsha. R. Olsson. Inductive functional programming using incremental program transformation. Artificial Intelligence, 74(1):55--83, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....append( c l , p, c q ) append( d y , 3,4 , d, e, f g ) append( h, z, i) append( x y , z, x u ) 39 append( 1,2 , 3,4 , 1,2,3,4 ) As the reader may notice, the above rule contains a lot of unnecessary literals. This problem is also discussed in [MF90] we is stated that the clauses that are induced by generalization can contain an extremely large number of (redundant) literals. If we have a background model M and n examples then the maximum number of literals in the created clause can be M n 1. A su#ciently large background model and or ....

....(literals without variables in general) from the clause, a few other methods also have been proposed for reducing the number of literals in the induced clause. 4.3.2. 6 Reducing the size of the clauses The first method for limiting the number of literals in the clauses is ijdetermination [MF90]. The idea behind this technique is to put a (weak) limitation on the hypothesis language: the clauses that are inducible are limited in the way that a maximum depth and degree is set on the variables that appear in the clause. This limitation prevents a combinatorial explosion in the number of ....

[Article contains additional citation context not shown here]

S. Muggleton and C. Feng. E#cient induction of logic programs. In First Conference on Algorithmic Learning Theory, 1990.

....by the ILP system. To begin with, we defined two relations, with plans of implementing full Allen s interval logic [All83] The two relations were during(EventId1, EventId2) and after(EventID1, EventID2) Unfortunately, this approach did not work well for a number of reasons. FOIL [Qui90] GOLEM [MF92] and PROGOL [Mug95] were all tested on simplified datasets with unsuccessful results. Each did not succeed for di#erent reasons. The first was the size of the dataset. Even with a small number of events and training instances, the number of generated predicates was huge especially that ....

S. Muggleton and C. Feng. E#cient induction of logic programs, 1992.

....from their discrete domains in a straightforward manner without prior partitioning. Usually, for symbolic data an explicit ordering scheme is given, like rule inference, which can be represented as data driven decision tree with labeled leaves [1, 2] Classic programs like FOIL [3] and GOLEM [4] both provide an induction of Horn clauses from data, but the domain of ILP has been extended to dynamic hypothesis generation [5] learning recursive logic [6] and program synthesis [7, 8] Soft representations are especially suitable for real value data, because the natural order of numbers ....

S. Muggleton and C. Feng. E#cient induction in logic programs. In S. Muggleton (ed.), Inductive Logic Programming, pp. 281--298, Academic Press, 1992.

....output is represented in two forms: left) graphical depiction and (right) English statement of the pattern. Figure courtesy A. Srinivasan (Oxford University) be attributed to a further restriction of the hypothesis space that permits certain optimizations) Some commercial systems (like Golem [Muggleton and Feng, 1990]) further require that background knowledge be ground, meaning that only base facts can be provided as opposed to intensional information (rules) This renders the overall complexity polynomial in the space of the database but pseudo polynomial (and sometimes exponential) in the space of the ....

Muggleton, S. and Feng, C. (1990). E#cient Induction of Logic Programs. In Arikawa, S., Goto, S., Ohsuga, S., and Yokomori, T., editors, Proceedings of the First International Conference on Algorithmic Learning Theory, pages 368--381. Japanese Society for Artificial Intelligence, Tokyo.

....While it is widely accepted that background knowledge is necessary for all but the simplest learning tasks, we note that there are two principle ways in which background knowledge can be used: 1. To expand the hypothesis language byintroducing extra terms #e.g. in Foil #Quinlan, 1990# and Golem #Muggleton and Feng, 1990##. 2. To constrain search #our objective in this paper#. These two methods have signi#cantly di#erent outcomes: in the #rst case background knowledge actually aggravates the search problem as the search space is expanded, whereas in the second the hypothesis space is restricted, reducing search. ....

Muggleton, S. and Feng, C. #1990#. E#cient induction of logic programs. In First International ConferenceonAlgorithmic Learning Theory, pages 369#381, Tokyo, Japan.

.... for the mode child( Determinate predicates are especially easy to evaluate, since no backtracking points need to be set, and in ILP systems, determinate predicates are often treated specially; for example, FOIL6.4 has special mechanisms for handling determinate predicates, and GOLEM [ Muggleton and Feng, 1992 ] allows only determinate predicates. Formal results also suggest that clauses containing determinate predicates are especially easy to learn [ Dzeroski et al. 1992; Cohen, 1995c ] Unfortunately, many of the predicates used in this problem are highly non determinate; in particular, the ....

....of such a clause will probably require a large amount of backtracking to evaluate, and hence will be expensive to evaluate. By appropriately restricting M , a user can avoid expensive clauses, without having to commit to any particular syntactic restriction on clauses such as determinacy [ Muggleton and Feng, 1992 ] or locality [ Cohen, 1994b ] Learning system Errors vs. default Time (sec) options) # W L p avg max # F1 FLIPPER 42 23.6 28 12 0.99 495 1,317 # F2 (monotone) 40 22.5 29 11 0.99 1,773 7,714 (monotone,modes) 49 27.5 22 12 0.87 3,944 24,439 M1 (M=100,monotone,modes) 63 35.4 21 26 0.46 384 ....

Stephen Muggleton and Cao Feng. E#cient induction of logic programs. In Inductive Logic Programming. Academic Press, 1992.

....process, and filters transform this format (on the fly) to the format required by the various data mining tools integrated into PYTHIA II. The goal is to accumulate tools that generate knowledge in the form of logic rules, if then else rules or decision trees. PYTHIA II first used GOLEM [Muggleton and Feng 1990], an empirical single predicate Inductive Logic Programming (ILP) learning system. It is a batch system that implements the relative least general generalization principle. We have experimented with other learning methods, e.g. fuzzy logic or neural networks, and have not found large di#erences ....

Muggleton, S. and Feng, C. 1990. E#cient induction of logic programs. In S. Arikawa, S. Goto, S. Ohsuga, and T. Yokomori (Eds.), Proceedings of the First International Conference on Algorithmic Learning Theory, pp. 368--381. Japanese Society for Artificial Intelligence, Tokyo.

....changes to the automatically generated knowledge as possible, we have integrated mostly systems that generate comprehensible knowledge in the form of logic rules, if then else rules or decision trees. The first learning system we integrated (we present some results using it later on) was GOLEM [Muggleton and Feng 1990], which is classified in [Dzeroski 1996] as an empirical single predicate Inductive Logic Programming (ILP) learning system. It is a batch system with noise handling capabilities that implements the relative least general generalization principle that can be considered as careful generalization in ....

Muggleton, S. and Feng, C. 1990. E#cient induction of logic programs. In S. Arikawa, S. Goto, S. Ohsuga, and T. Yokomori (Eds.), Proceedings of the First International Conference on Algorithmic Learning Theory, pp. 368--381. Japanese Society for Artificial Intelligence, Tokyo.

.... explicit background knowledge to the system in the form of additional relations or axioms 1 A determinate literal is one that introduces new variables so that there is exactly one binding for each positive example and at most one binding for each negative example in the partially constructed rule (Muggleton Feng, 1992; Quinlan, 1996) Table 3. A simple logistics problem and its solution. There are three cities (A, B, and C) each containing two locations, an airport and a posto#ce (e.g. apt A and po A) In each city there is a truck (trk A, trk B, trk C) and there is one airplane (pln) There are two ....

Muggleton, S., & Feng, C. (1992). E#cient induction of logic programs. In S. Muggleton(Ed.), Inductive Logic Programming . London: Academic Press Limited.

....natural and intuitive technique to use for learning from entailment, and it has been used # This work was partly supported by EPSRC Grant GR M21409. 1 The unknown expression that has to be identified is commonly referred to as target expression. 1 before, both in theoretical and applied work [Ari97, RT98, RS98, MF92]. Thus the contributions of this paper are to give a more direct algorithm for the class and establish better bounds in terms of running time and number of queries to the oracles. We extend our results to the class of fully inequated range restricted Horn expressions. The main property of this ....

S. Muggleton and C. Feng. E#cient induction in logic programs. In S. Muggleton, editor, Inductive Logic Programming, pages 281--298. Academic Press, 1992.

....as subterms of more complex terms. And every clause includes in its antecedent all inequalities possible between all terms appearing in it. The paper shows that small modifications to the algorithm and proof of [AK00] yield the learning result. Further background and related work appear in [Ari97, RT98, RS98, MF92, Kha99a, Kha99b]. The rest of the paper is organised as follows. Section 2 gives some preliminary definitions. The learning algorithm is presented in Section 3 and proved correct in Section 4. 2 Preliminaries 2.1 Inequated Range Restricted Horn Expressions We consider a subset of the class of universally ....

S. Muggleton and C. Feng. E#cient induction in logic programs. In S. Muggleton, editor, Inductive Logic Programming, pages 281--298. Academic Press, 1992.

....generalizations, for instance by Plotkin #1971b#, Muggleton and Page #1994#, Idestam Almquist #1993, 1995#, Niblett #1988#, though not b y Plotkin #1970#, but we feel this general is redundant. 4. There is also a relation between least generalization under subsumption and inverse resolution #Muggleton, 1992#. 2 Least Generalizations and Greatest Specializations Because of the weakness of subsumption, it is desirable to make the step from the subsumption order to the more powerful implication order. Accordingly, it is importantto #nd out whether Plotkin s positive result on the existence of LGS s ....

....is rather restricted: it must be a #nite set # of ground literals. Because of its restrictiveness, wehave not included relative subsumption in Table 1. Nevertheless, we mention it here, because least generalization under relative subsumption forms the basis of the well known ILP system Golem #Muggleton Feng, 1992#. De#nition 13 Let C;D be clauses, # =fL 1 ; L m g be a #nite set of ground literals. Then C subsumes D relative to #, denoted by C ## D,ifC # #D #f:L 1 ; L m g#. 2 10 Least Generalizations and Greatest Specializations It is easy to see that ## is re#exive and transitive, so it imposes ....

Muggleton, S., & F eng, C. #1992#. E#cient induction of logic programs. In Muggleton, S.

....changes to the automatically generated knowledge as possible# we have integrated mostly systems that generate comprehensible knowledge in the form of logic rules# if#then#else rules or decision trees. The #rst learning system we integrated #we present some results using it later on## was GOLEM #Muggleton and Feng 1990## which is classi#ed in #Dzeroski 1996# as an empirical single predicate Inductive Logic Programming #ILP# learning system. It is a batch non#interactive system with noise handling capabilities that implements # 19 the relative least general generalization principle that can be considered as ....

Muggleton# S. and Feng# C. 1990. E#cient induction of logic programs. In S. Arikawa# S. Goto# S. Ohsuga# and T. Yokomori #Eds.## Proceedings of the First International Conference on Algorithmic Learning Theory# pp. 368#381. Japanese Society for Arti#cial Intelligence# Tokyo.

No context found.

S. H. Muggleton and C. Feng. E#cient induction of logic programs. In Proceedings of the First Conference on Algorithmic Learning Theory, pages 368--381, Tokyo, 1990. Ohmsha. R. Olsson. Inductive functional programming using incremental program transformation. Artificial Intelligence, 74(1):55--83, 1995.

No context found.

S. Muggleton & C. Feng. E#cient induction of logic programs. Proc. First Conf. on Algorithmic Learning Theory, Ohmsha, Tokyo, 1990. Also in [30], pp.281--298.

No context found.

S. Muggleton, C. Feng, E#cient induction of logic programs, in: First Conference on Algorithmic Learning Theory, 1990.

No context found.

S. H. Muggleton and C. Feng. E#cient induction of logic programs. In S. H. Muggleton, editor, Inductive Logic Programming, pages 281--298. Acadamic Press, 1992.

No context found.

S. Muggleton and C. Feng. E#cient induction in logic programs. In S. Muggleton, editor, Inductive Logic Programming, pages 281--298. Academic Press, 1992.

No context found.

S. Muggleton and C. Feng. E#cient induction in logic programs. In S. Muggleton, editor, Inductive Logic Programming, pages 281--298. Academic Press, 1992.

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