Luc Dehaspe. 1997. Maximum entropy modeling with clausal constraints. In Proceedings of the 7th International Workshop on Inductive Logic Programming.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....ILP systems, and thus can be said to be a successful upgrade. 6 Related Work and Conclusions There are plenty of other inductive logic programming systems whose development more or less ts in with the proposed methodology: Foil [58] RIBL [36] SRT [46] Tilde [9, 7] Warmr [26, 25] Maccent [24], jk CT learner [21] Claudien [20] Probabilistic Relational Models [45] Cohen s Flipper [16] 61] and RDBC [44] e.g. Quinlan s Foil can also be considered an upgrade of either Michalski s AQ [47] or CN2 [14, 13] RIBL upgrades the classical k nearest neighbor algorithm (using a rst order ....

L. Dehaspe. Maximum entropy modeling with clausal constraints. In Proceedings of the Seventh International Workshop on Inductive Logic Programming, volume 1297 of Lecture Notes in Arti cial Intelligence, pages 109-124. Springer-Verlag, 1997.

.... important lesson learned during the development of several inductive logic programming systems and results of the machine learning group in Leuven (including [De Raedt and D eroski, 1994] TLDE [Blockeel and De Raedt, 1998; Blockeel, 1998] ICL [De Raedt and Van Laer, 1995] CLAUDraN [De Raedt and Dehaspe, 1997a] and WARMR [Dehaspe and De Raedt, 1997; Dehaspe, 1998] of which some are briefly discussed in Section 4.10) The method starts from an existing propos itional learner and provides a recipe for upgrading it towards the use of first order logic. The recipe involves the use of examples which ....

.... inverse implication, inverse resolution and inverse entailment (see [Muggleton and De Raedt, 1994] for an overview) However, in practice, the large majority of ILP systems (including FOIL [Quinlan, 1990] GOLEM [Muggleton and Feng, 1990] PROGOL TM [Muggleton, 1995] CLAUDIEN [De Raedt and Dehaspe, 1997a] and TILDE [Blockeel and De Raedt, 1998; Blockeel, 1998] uses 0 subsumption. This is due to the better computational properties of 0 subsumption as compared to inverse resolution and inverse implication (which are both computationally intractable and less well understood) The fact that ....

[Article contains additional citation context not shown here]

L. Dehaspe. Maximum entropy modeling with clausal con- straints. In Proceedings of the Seventh International Workshop on Inductive Logic Programming, volume 1297 of Lecture Notes in Artificial Intelligence, pages 109-124. Springer-Verlag, 1997.

....ILP systems, and thus can be said to be a successful upgrade. 6 Related Work and Conclusions There are plenty of other inductive logic programming systems whose development more or less fits in with the proposed methodology: Foil [57] RIBL [35] SRT [45] Tilde [9, 7] Warmr [26, 25] Maccent [24], jk CT learner [21] Claudien [20] Probabilistic Relational Models [44] Cohen s Flipper (in [17] 60] and RDBC [43] e.g. Quinlan s Foil can also be considered an upgrade of either Michalski s AQ (1983) or CN2, RIBL upgrades the classical k nearest neighbor algorithm (using a first order ....

L. Dehaspe. Maximum entropy modeling with clausal constraints. In Proceedings of the Seventh International Workshop on Inductive Logic Programming, pages 109--124. Springer-Verlag, 1997. 25

....logical decision trees (binary trees are a special case) As a result, the hypothesis space is different from ICL, and a different language specification than DLAB is used. Two more examples are Warmr [12] which extends Apriori [1] to mine association rules in multiple relations, and Maccent [11], which is an upgrade of the maximum entropy approach in [2] 4 Conclusions Learning from interpretations is a nice framework for upgrading a propositional learning system towards a first order learning system. We have formulated a methodology for doing such an upgrade and have evaluated this ....

L. Dehaspe. Maximum entropy modeling with clausal constraints. In Proceedings of the 7th International Workshop on Inductive Logic Programming, volume 1297 of Lecture Notes in Artificial Intelligence, pages 109-- 124. Springer-Verlag, 1997.

....is defined as the difference between the entropy of the a priori class distribution and the conditional entropy of the classes given the value of the feature. 6 2. 4 Maximum Entropy combination The second machine learning method, Maximum Entropy Modeling, implemented in the MACCENT system (Dehaspe, 1997), 7 does the classification task by selecting the most probable class given a Maximum Entropy Model. This type of model represents examples of the task (Cases) as sets of binary indicator features, for the task at hand conjunctions of a particular tag and a particular set of feature values. The ....

Dehaspe, L. 1997. Maximum entropy modeling with clausal constraints. In Inductive Logic Programming: Proceedings of the 7th International Workshop (ILP-97), Lecture Notes in Artificial Intelligence, 1297, pages 109--124. Springer Verlag.

....Srinivasan and King [13] used Progol to construct relationally defined boolean features and then used these in linear regression. Pompe and Kononenko [10] used ILP R to build an hypothesis and then used a Bayesian approach involving splitting and merging clauses to do classification. Dehaspe [5] explicitly constructs a conditional probability distribution using a maximum entropy approach with clausal constraints. In all these cases, experimental results demonstrate the effectiveness of a probabilistic approach. In this paper we use a particular application of Bayes theorem which allows ....

Luc Dehaspe. Maximum entropy modeling with clausal constraints. In Inductive Logic Programming: Proceedings of the 7th International Workshop (ILP-97). LNAI 1297, pages 109--124. Springer, 1997.

....and where there already exists valuable work c 2000 Kluwer Academic Publishers. Printed in the Netherlands. submit.tex; 31 03 2000; 11:36; p. 2 3 involving probabilistic methods (Pompe and Kononenko, 1995; Muggleton, 1996; Pompe and Kononenko, 1997; Flach and Lachiche, 1999) Most notable is (Dehaspe, 1997) where, in work related to the current paper, Dehaspe combines maximum entropy and ILP methods in the MACCENT algorithm. The connections and contrasts between Dehaspe s log linear models with clausal constraints and SLPs can be found in (Cussens, 1999b) Despite this existing work, it is fair to ....

....models that represent uncertainty. This paper is lop sided in the opposite direction, focusing exclusively on a parametric statistical analysis of SLPs, so that a large number of topics examined in related work are expressly left out. Most noticeably, we do not examine structure learning as (Dehaspe, 1997) did. Also we make no attempt to connect the semantics of SLPs to that of logic programs as in (Muggleton, 2000) There is also little on the connections between SLPs and related approaches, which is discussed in (Cussens, 1999a; Cussens, 1999b) The paper is organised as follows. First note ....

[Article contains additional citation context not shown here]

Dehaspe, L.: 1997, `Maximum Entropy Modeling with Clausal Constraints'. In: Inductive Logic Programming: Proceedings of the 7th International Workshop (ILP-97). LNAI 1297. pp. 109-124.

....features of our distribution, it is natural to look to ILP for techniques which induce such structural features from data. Work in ILP on learning from positive examples only [15, 3] is of relevance here, but the most thorough incorporation of probabilistic approaches into ILP is by Dehaspe in [5]. Dehaspe presents the MACCENT algorithm which constructs a log linear model using boolean clausal constraints as features. Dehaspe uses the learning from interpretations ILP setting where each example is a Prolog database. This approach to ILP gathers information relevant to a particular ....

Luc Dehaspe. Maximum entropy modeling with clausal constraints. In Inductive Logic Programming: Proceedings of the 7th International Workshop (ILP-97). LNAI 1297, pages 109--124. Springer, 1997.

....policy if many of these indicators point in the right direction. If some of the feature values point in the opposite direction, it becomes less likely that they will be interested. For the Benelearn competition a large variety of algorithms were tried: neural networks, maximum entropy modelling [5], Naive Bayes, decision trees, rule based approaches (both rules derived from trees and rules induced using a genetic algorithm) linear regression, logistic regression, Approaches that scored very well in this competition were Gentle AdaBoost (based on logistic regression) Maccent, Naive ....

....features Most ILP algorithms (e.g. Progol [17] FOIL [19] ICL [4] are essentially rule set induction systems, and hence are biased towards non cumulativity. Some exceptions are tree based [1, 15] instance based [9] and probabilistic systems (e.g. Bayesian learners [10] Maccent [5], or Cussens probabilistic approaches [3] None of these seem to be as widely used as FOIL and Progol though. Possibly, in domains where cumulativity is important, such systems (and especially the probabilistic ones) deserve more attention than given up till now. The approaches mentioned above ....

L. Dehaspe. Maximum entropy modeling with clausal constraints. In Proceedings of the Seventh International Workshop on Inductive Logic Programming, volume 1297 of Lecture Notes in Artificial Intelligence, pages 109--124. Springer-Verlag, 1997.

No context found.

Luc Dehaspe. 1997. Maximum entropy modeling with clausal constraints. In Proceedings of the 7th International Workshop on Inductive Logic Programming.

No context found.

Luc Dehaspe. 1997. Maximum entropy modeling with clausal constraints. In Proceedings of the 7th International Workshop on Inductive Logic Programming.

No context found.

L. Dehaspe. Maximum entropy modeling with clausal constraints. In ILP, 1997.

No context found.

Luc Dehaspe. 1997. Maximum entropy modeling with clausal constraints. In Proceedings of the 7th International Workshop on Inductive Logic Programming.

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