## A Perspective on Inductive Logic Programming

### Cached

### Download Links

Citations: | 56 - 8 self |

### BibTeX

@MISC{Raedt_aperspective,

author = {Luc De Raedt},

title = {A Perspective on Inductive Logic Programming},

year = {}

}

### Years of Citing Articles

### OpenURL

### Abstract

. The state-of-the-art in inductive logic programming is surveyed by analyzing the approach taken by this field over the past 8 years. The analysis investigates the roles of 1) logic programming and machine learning, of 2) theory, techniques and applications, of 3) various technical problems addressed within inductive logic programming. 1 Introduction The term inductive logic programming was first coined by Stephen Muggleton in 1990 [1]. Inductive logic programming is concerned with the study of inductive machine learning within the representations offered by computational logic. Since 1991, annual international workshops have been organized [2-8]. This paper is an attempt to analyze the developments within this field. Particular attention is devoted to the relation between inductive logic programming and its neighboring fields such as machine learning, computational logic and data mining, and to the role that theory, techniques and implementations, and applications play. The analysis...

### Citations

8089 | Maximum likelihood from incomplete data via the EM algorithm
- Dempster, Laird, et al.
- 1977
(Show Context)
Citation Context ...rical optimization problem, and all known algorithms involve nonlinear optimization The most commonly adapted technique for probabilistic logic learning is the Expectation-Maximization (EM) algorithm =-=[10,28]-=-. EM is based on the observation that learning would be easy (i.e., correspond to frequency counting), if the values of all the random variables would be known. Therefore, it first estimates these val... |

7052 |
Probabilistic Reasoning in Intelligent Systems
- Pearl
- 1988
(Show Context)
Citation Context ...roaches have been developed such as [38]. In this paper, we choose Bayesian logic programs [21] as the probabilistic logic programming system because Bayesian logic programs combine Bayesian networks =-=[37]-=-, which represent probability distributions over propositional interpretations, with definite clause logic. Furthermore, Bayesian logic programs have already been employed for learning. The idea under... |

2105 | Building a Large Annotated Corpus of English: The Penn Treebank
- Marcus, Marcinkiewicz, et al.
- 1993
(Show Context)
Citation Context ...provide this type of examples. Even though that is generally true, there exist specific situations for which this is feasible. Indeed, consider tree banks such as the UPenn Wall Street Journal corpus =-=[27]-=-, which contain parse trees. These trees directly correspond to the proof-trees we talk about. Even {}sthough – to the best of the authors’ knowledge (but see [43,3] for inductive logic programming sy... |

1694 | A Theory of the Learnable
- Valiant
- 1984
(Show Context)
Citation Context ...r under θ-subsumption [39]. 2.2 Learning from Interpretations The learning from interpretations settings [8] has been inspired on the work on boolean concept-learning in computational learning theory =-=[47]-=-.sDefinition 2. When learning from interpretations, the examples are Herbrand interpretations and a hypothesis H covers an example e w.r.t. the background theory B if and only if e is a model of B ∪ H... |

1055 | Inductive logic programming
- Muggleton
- 1991
(Show Context)
Citation Context ...erent types of probabilistic representations. 1 Introduction In the past few years there has been a lot of work lying at the intersection of probability theory, logic programming and machine learning =-=[40,15,42,31,35, 18,25,21,2,24]-=-. This work is known under the names of statistical relational learning [14,11], probabilistic logic learning [9], or probabilistic inductive logic programming. Whereas most of the existing works have... |

854 | A tutorial on learning with bayesian networks
- Heckerman
- 1995
(Show Context)
Citation Context ...theory revision approaches in inductive logic prorgamming. In case that only propositional clauses are considered, Scooby coincides with greedy hill-climbing approaches for learning Bayesian networks =-=[16]-=-. 4.5 Learning from Probabilistic Proofs Given a training set E containing ground proofs as examples, one possible approach combines ideas from the early inductive logic programming system Golem [34] ... |

853 | Leaming Logical Definitions from Relations - Quinlan - 1990 |

768 | Constraint Logic Programming: A Survey - Jaffar, Maher - 1994 |

627 | Inverse entailment and progol
- Muggleton
- 1995
(Show Context)
Citation Context ...arning from entailment is by far the most popular inductive logic programming systems and it is addressed by a wide variety of well-known inductive logic programming systems such as FOIL [41], Progol =-=[30]-=-, and Aleph [44]. Definition 1. When learning from entailment, the examples are definite clauses and a hypothesis H covers an example e w.r.t. the background theory B if and only if B ∪ H |= e. In man... |

596 | An Introduction to Computational Learning Theory - Kearns, Vazirani - 1994 |

572 |
Bayesian Networks and Decision Graphs
- Jensen
- 2002
(Show Context)
Citation Context ...on cpd(c) = P(A | A1, . . .,An), which quantifies as a macro the probabilistic dependency among ground instances of the clause. Let us illustrate Bayesian logic programs on Jensen’s stud farm example =-=[19]-=-, which describes the processes underlying a life threatening heridatary disease. Example 5. Consider the following Bayesian clauses: carrier(X) | founder(X). (3) carrier(X) | mother(M,X), carrier(M),... |

470 |
Algorithmic Program DeBugging
- Shapiro
- 1983
(Show Context)
Citation Context ..., most notably learning from entailment [39] and learning from interpretations [8,17], which we formalize below. We also introduce an intermediate setting inspired on the seminal work by Ehud Shapiro =-=[43]-=-, which we call learning from proofs. Before formalizing these settings, let us however also discuss how background knowledge is employed within inductive logic programming. For the purposes of the pr... |

463 |
A note on inductive generalization
- Plotkin
- 1970
(Show Context)
Citation Context ...g the examples together with the covers relation determines the inductive logic programming setting [6]. Various settings have been considered in the literature, most notably learning from entailment =-=[39]-=- and learning from interpretations [8,17], which we formalize below. We also introduce an intermediate setting inspired on the seminal work by Ehud Shapiro [43], which we call learning from proofs. Be... |

455 | Inductive logic programming: Theory and methods - Muggleton, Raedt - 1994 |

294 | Probabilistic Horn abduction and Bayesian networks
- Poole
- 1993
(Show Context)
Citation Context ...erent types of probabilistic representations. 1 Introduction In the past few years there has been a lot of work lying at the intersection of probability theory, logic programming and machine learning =-=[40,15,42,31,35, 18,25,21,2,24]-=-. This work is known under the names of statistical relational learning [14,11], probabilistic logic learning [9], or probabilistic inductive logic programming. Whereas most of the existing works have... |

288 | The Penn Treebank: Annotating predicate– argument structure
- Marcus, Kim, et al.
- 1994
(Show Context)
Citation Context ...provide this type of examples. Even though that is generally true, there exist specific situations for which this is feasible. Indeed, consider tree banks such as the UPenn Wall Street Journal corpus =-=[26]-=-, which contain parse trees. These trees directly correspond to the proof-trees we talk about. Even {}sthough – to the best of the authors’ knowledge (but see [42, 3] for inductive logic programming s... |

188 | Probabilistic frame-based systems
- Koller, Pfeffer
- 1998
(Show Context)
Citation Context ...erent types of probabilistic representations. 1 Introduction In the past few years there has been a lot of work lying at the intersection of probability theory, logic programming and machine learning =-=[40,15,42,31,35, 18,25,21,2,24]-=-. This work is known under the names of statistical relational learning [14,11], probabilistic logic learning [9], or probabilistic inductive logic programming. Whereas most of the existing works have... |

184 | A theory of clausal discovery
- Raedt, Bruynooghe
- 1993
(Show Context)
Citation Context ... from interpretations is well suited for learning from positive examples only. For this case, a complete search of the space ordered by θ-subsumption is performed until all clauses cover all examples =-=[7]-=-. 2.3 Learning from Proofs Because learning from entailment (with ground facts as examples) and interpretations occupy extreme positions w.r.t. the information the examples carry, it is interesting to... |

152 | Theories for mutagenicity: a study in first-order and feature-based induction
- Srinivasan, Muggleton, et al.
- 1996
(Show Context)
Citation Context ...ne requires that the examples are ground facts, a special case of definite clauses. To illustrate the above setting, consider the following example inspired on the well-known mutagenicity application =-=[45]-=-. Example 1. Consider the following facts in the background theory B. They describe part of molecule 225. molecule(225). bond(225,f1_1,f1_2,7). logmutag(225,0.64). bond(225,f1_2,f1_3,7). lumo(225,-1.7... |

144 | Stochastic attribute-value grammars
- Abney
- 1997
(Show Context)
Citation Context ...any Bayesian network inference engine based on Equation (6). 3.3 Probabilistic Proofs The framework of stochastic logic programs [12,31,4] was inspired on research on stochastic context free grammars =-=[1, 26]-=-. The analogy between context free grammars and logic programs is that 1. grammar rules correspond to definite clauses, 2. sentences (or strings) to atoms, and 3. derivations to proofs. Furthermore, i... |

136 | Separate-and-Conquer Rule Learning
- Fürnkranz
- 1999
(Show Context)
Citation Context ... the following clause mutagenic(M) :- nitro(M,R1), logp(M,C), C > 1 . Inductive logic programming systems that learn from entailment often employ a typical separate-and-conquer rule-learning strategy =-=[13]-=-. In an outer loop of the algorithm, they follow a set-covering approach [29] in which they repeatedly search for a rule covering many positive examples and none of the negative examples. They then de... |

134 | Hidden markov model induction by bayesian model merging
- Stolcke, Omohundro
- 1993
(Show Context)
Citation Context ...es ideas from the early inductive logic programming system Golem [33] that employs Plotkin’s [38] least general generalization (LGG) with bottom-up generalization of grammars and hidden Markov models =-=[45]-=-. The resulting algorithm employs the likelihood of the proofs score(L, λ, E) as the scoring function. It starts by taking as L0 the set of ground clauses that have been used in the proofs in the trai... |

134 | An algorithm for multi-relational discovery of subgroups - Wrobel - 1997 |

129 | Inducing probabilistic grammars by Bayesian model merging
- Stolcke, Omohundro
- 1994
(Show Context)
Citation Context ...es ideas from the early inductive logic programming system Golem [34] that employs Plotkin’s [39] least general generalization (LGG) with bottom-up generalization of grammars and hidden Markov models =-=[46]-=-. The resulting algorithm employs the likelihood of the proofs score(L, λ, E) as the scoring function. It starts by taking as L0 the set of ground clauses that have been used in the proofs in the trai... |

109 | Bayesian logic programs
- Kersting, Raedt
- 2001
(Show Context)
Citation Context |

101 | Relational bayesian networks
- Jaeger
- 1997
(Show Context)
Citation Context |

95 | A statistical learning method for logic programs with distribution semantics
- Sato
- 1995
(Show Context)
Citation Context |

94 | Answering queries from context-sensitive probabilistic knowledge bases
- Ngo, Haddawy
- 1997
(Show Context)
Citation Context |

84 | Relational Markov models and their application to adaptive Web navigation
- Anderson, Domingos, et al.
(Show Context)
Citation Context |

76 | Towards combining inductive logic programming and Bayesian networks
- Kersting, Raedt
- 2001
(Show Context)
Citation Context ...ble formalisms and techniques, including probabilistic Horn abduction by David Poole, PRISMs by Sato, stochastic logic programs by Eisele [12], Muggleton [31] and Cussens [4], Bayesian logic programs =-=[22,20]-=- by Kersting and De Raedt, and Logical Hidden Markov Models [24]. The main contribution of this paper is the introduction of three probabilistic inductive logic programming settings which are derived ... |

71 | The application of inductive logic programming to finite element mesh design - Dolaak, Muggleton - 1992 |

69 | Parameter estimation in stochastic logic programs
- Cussens
- 2001
(Show Context)
Citation Context .... The first actually corresponds to a learning from entailment setting in which the examples are ground atoms entailed by the target stochastic logic program. This setting has been studied by Cussens =-=[5]-=-, who solves the parameter estimation problem, and Muggleton [32,33], who presents a preliminary approach to structure learning (concentrating on the problem of adding one clause to an existing stocha... |

61 | Probabilistic reasoning for complex systems
- Pfeffer
- 2000
(Show Context)
Citation Context ...obabilities to interpretations covered by an annotated logic program. In the past few years, this question has received a lot of attention and various different approaches have been developed such as =-=[38]-=-. In this paper, we choose Bayesian logic programs [21] as the probabilistic logic programming system because Bayesian logic programs combine Bayesian networks [37], which represent probability distri... |

61 | Induction of logic programs: Foil and related systems
- Quinlan, Cameron-Jones
- 1995
(Show Context)
Citation Context ...c representations, examples and probability distributions. The first setting, probabilistic learning from entailment, combines key principles of the well-known inductive logic programming system FOIL =-=[41]-=- with thesnaïve Bayes’ assumption; the second setting, probabilistic learning from interpretations, incorporated in Bayesian logic programs [22,20], integrates Bayesian networks with logic programming... |

55 |
Inductive Logic Programming From Machine Learning to Software Engineering
- Bergadano, Gunetti
- 1996
(Show Context)
Citation Context ...the UPenn Wall Street Journal corpus [27], which contain parse trees. These trees directly correspond to the proof-trees we talk about. Even {}sthough – to the best of the authors’ knowledge (but see =-=[43,3]-=- for inductive logic programming systems that learn from traces) – no inductive logic programming system has been developed to learn from proof-trees, it is not hard to imagine an outline for such an ... |

52 | Generating Bayesian networks from probability logic knowledge bases
- Haddawy
- 1994
(Show Context)
Citation Context |

42 |
Induction as nonmonotonic inference
- Helft
- 1989
(Show Context)
Citation Context ... relation determines the inductive logic programming setting [6]. Various settings have been considered in the literature, most notably learning from entailment [39] and learning from interpretations =-=[8,17]-=-, which we formalize below. We also introduce an intermediate setting inspired on the seminal work by Ehud Shapiro [43], which we call learning from proofs. Before formalizing these settings, let us h... |

41 |
The EM Algorithm and Extensions
- McLachan, Krishnan
- 1997
(Show Context)
Citation Context ...rical optimization problem, and all known algorithms involve nonlinear optimization The most commonly adapted technique for probabilistic logic learning is the Expectation-Maximization (EM) algorithm =-=[10,28]-=-. EM is based on the observation that learning would be easy (i.e., correspond to frequency counting), if the values of all the random variables would be known. Therefore, it first estimates these val... |

41 | Foundations of Inductive Logic Programming - Nienhuys-Cheng, Wolf - 1997 |

38 | Relational instance based learning - Emde, Wettschereck - 1996 |

35 | Loglinear models for first-order probabilistic reasoning
- Cussens
- 1999
(Show Context)
Citation Context ...buted a rich variety of valuable formalisms and techniques, including probabilistic Horn abduction by David Poole, PRISMs by Sato, stochastic logic programs by Eisele [12], Muggleton [31] and Cussens =-=[4]-=-, Bayesian logic programs [22,20] by Kersting and De Raedt, and Logical Hidden Markov Models [24]. The main contribution of this paper is the introduction of three probabilistic inductive logic progra... |

34 | Probabilistic Logic Learning
- Raedt, Kersting
(Show Context)
Citation Context ...probability theory, logic programming and machine learning [40,15,42,31,35, 18,25,21,2,24]. This work is known under the names of statistical relational learning [14,11], probabilistic logic learning =-=[9]-=-, or probabilistic inductive logic programming. Whereas most of the existing works have started from a probabilistic learning perspective and extended probabilistic formalisms with relational aspects,... |

32 | Relational distance-based clustering - Kirsten, Wrobel - 1998 |

31 | Top-down induction of logical decision trees - Blockeel, Raedt - 1998 |

29 |
Learning Structure and Parameters of Stochastic Logic Programs
- Muggleton
- 2002
(Show Context)
Citation Context ...etting in which the examples are ground atoms entailed by the target stochastic logic program. This setting has been studied by Cussens [5], who solves the parameter estimation problem, and Muggleton =-=[32,33]-=-, who presents a preliminary approach to structure learning (concentrating on the problem of adding one clause to an existing stochastic logic program). In the second setting, which we sketch below, t... |

27 |
First-order jk-clausal theories are PAC-learnable
- Raedt, D˘zeroski
- 1994
(Show Context)
Citation Context ... relation determines the inductive logic programming setting [6]. Various settings have been considered in the literature, most notably learning from entailment [39] and learning from interpretations =-=[8,17]-=-, which we formalize below. We also introduce an intermediate setting inspired on the seminal work by Ehud Shapiro [43], which we call learning from proofs. Before formalizing these settings, let us h... |

24 | Adaptive Bayesian logic programs
- Kersting, Raedt
- 2001
(Show Context)
Citation Context ...ble formalisms and techniques, including probabilistic Horn abduction by David Poole, PRISMs by Sato, stochastic logic programs by Eisele [12], Muggleton [31] and Cussens [4], Bayesian logic programs =-=[22,20]-=- by Kersting and De Raedt, and Logical Hidden Markov Models [24]. The main contribution of this paper is the introduction of three probabilistic inductive logic programming settings which are derived ... |

24 | Polynomial learnability and Inductive Logic Programming: methods and results - Cohen, Page - 1995 |

18 |
Raedt. Logical settings for concept-learning
- De
- 1997
(Show Context)
Citation Context ...three probabilistic inductive logic programming settings which are derived from the learning from entailment, from interpretations and from proofs settings of the field of inductive logic programming =-=[6]-=-. Each of these settings contributes different notions of probabilistic logic representations, examples and probability distributions. The first setting, probabilistic learning from entailment, combin... |

18 |
Raedt. Towards discovering structural signatures of protein folds based on logical hidden markov models
- Kersting, Raiko, et al.
- 2003
(Show Context)
Citation Context |