H. Blockeel, L. De Raedt, and J. Ramon. Top-down induction of logical decision trees. In Proc. 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....logic programming, belief nets and link analysis. A number of approaches extending one table learners to multi table domains have been proposed in the inductive logic programming (ILP) community. Generally, these approaches extend learners most suitable to purely binary attributes. Tilde [Blockeel and Raedt, 1998] and WARMR [Dehaspe and Toivonen, 1999] for example, extend decision trees and association rules, respectively. Another ILP approach is propositionalization. It uses bodies of first order rules learned by an ILP technique as binary features in attribute value learners. Kramer et al. 2001] ....

Hendrik Blockeel and Luc De Raedt. Top-down induction of logical decision trees. Artificial Intelligence, 101(1-2):285--297, 1998.

....those used in KDD Cup 2001. The idea of shifting from ILP techniques to a search space consisting of database queries using SQL instead of logical clauses as the language bias and results from database theory as the next step in the relational data mining field had been already proposed in (Blockeel and De Raedt, 1997a) There, Blockeel et al. borrow techniques from ILP to tackle the problem of finding relationships between relations using relational algebra as the language bias. An equivalent algorithm using SQL is straightforward. That algorithm has not been implemented yet. Chapter 2 discusses some of the ....

....easy to use, can be used by a range of ILP systems, and can represent the complexity of the problem in a clear and simple way. In that way, even the non expert users will be able to model their problems and use a wide range of ILP engines, choosing that one that best fits their current needs. In (Blockeel and De Raedt, 1997a) three ways of bridging ILP and relational databases are presented. These are possible solutions to the problem of efficiency issues mentioned before. They are briefly described in order of adaptation of the ILP system to the database system. The simplest one is pre processing of the ....

[Article contains additional citation context not shown here]

Blockeel, H., and De Raedt, L. Top-down induction of Logical Decision Trees. In W. K. Van Mareke, editor, Proceedings of the Ninth Dutch Conference on Artificial Intelligence (NAIC'97), 1997.

.... and inconsistent theories that satisfy some other acceptance criteria (predictive accuracy, significance, compression) the predictive ILP setting can be extended to include learning of classification and prediction rules from imperfect data, as well as learning of logical decision trees [2]. In a broader sense, predictive ILP incorporates also first order regression [33] and constraint inductive logic programming [53] for which again different acceptance criteria apply. 2.2.2 Descriptive ILP Descriptive ILP is sometimes referred to as confirmatory induction, nonmonotonic ILP, ....

....and negative examples and background knowledge; in addition, Progol can also be used to learn from positive examples only. They use different acceptance criteria: compression, coverage accuracy and minimal description length, respectively. Induction of logical decision trees. The system Tilde [2] belongs to Top down induction of decision tree algorithms. It can be viewed as a firstorder upgrade of Quinlan s C4.5, employing logical queries in tree nodes which involves appropriate handling of variables. The main advantage of Tilde is its efficiency and capability of dealing with large ....

[Article contains additional citation context not shown here]

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Submitted to DAMI, Special Issue on Inductive Logic Programming, 1998.

....= 5000, balance = 2 , exit = false and splits = 1. We ran a 10 fold cross validation (taking about 12 hours) with these settings and found the average accuracy was 83.46 with a standard deviation of 9.51 . This accuracy is similar to that obtained by another general purpose decision tree learner [BD97, BD98], but lower than the 92.4 accuracy obtained by a learning system specially developed for this problem and described in [DLLP97] and also lower than the 88.9 accuracy in [MLP98] obtained by probabilistic methods specially developed for multiple instance problems. With the same settings, the ....

....( 1) 2) In this context, the learner searched for a definition for mutagenic with the settings prune = 75 , cutout = 400, balance = 2 , exit = false and splits = 0. On a 10 fold cross validation (taking about 5. 5 hours) the average accuracy was 87.3 (very similar to the results given in [SMKS94, KMSS96, SKM99, BD97, BD98]) with a standard deviation of 6.10 . The learner found the following definition on the full data set. In fact, this same theory was also produced for each of the 10 training sets in the cross validation. mutagenic m = if and2 (projLumo. 2.368) projInd1. False) m then False else ....

[Article contains additional citation context not shown here]

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Available at http://www.cs.kuleuven.ac.be/~hendrik, 1997.

....of iterations and transforming the resulting ensemble into a single decision tree (maybe similar to work described in [Pfahringer 97] might be an interesting starting point for such a synthesis. Furthermore, we plan to adapt the key ideas of PreC4 to learning in a first order framework (like [Blockeel DeRaedt 97] or [Kramer 96] where due to vastly larger concept spaces effective pruning is even more essential than it already is in a propositional setting. Acknowledgements The first idea for PreC4 was conceived while the author was visiting the Computer Science Department of the University of Waikato, ....

Blockeel H., DeRaedt L.: Top-down induction of logical decision trees, Technical Report Report CW 247, Katholieke Universiteit Leuven, Belgium, 1997.

....represent nominal values, a set of binary monitors can be used to represent all possible values of the original monitor. Monitors that have numeric values, such as the response time to a request, can be discretised first. First order Inductive logic programming Inductive logic programming (ILP) [4 , 11, 12, 13, 15] offers a knowledge rich approach to machine learning: ILP can incorporate domain knowledge into the modelling process in the form of a logic program which specifies the links and relationships already known to be in the domain of application, that is the background knowledge. In the domain of ....

....background knowledge which allows learned clauses to be incorporated for use in later learning. It constructs a pruned top down search which produces guaranteed optimally short clauses. Furthermore, it supports numerical hypotheses using built in functions as background knowledge. Tilde Tilde [4] is a recent ILP system that builds binary decision trees, having no overlapping coverage. It is derived from Quinlan s C4.5, in that Tilde uses the same heuristics and search strategy. However, it works with a first order representation. Hence, it has C4.5 (when working with binary attributes) as ....

Blockeel, H., De Raedt, L. Top-down induction of logical decision trees, submitted to Artificial Intelligence, 1997.

....where the A i are atoms 1 . Typically, n = 1 and the specification will be of the form rmode(n : atom) This 1 Alternatively, Warmr s language bias can be specified in Dlab format [19] as in Claudien [16] and ICL [18] format, originally proposed for Progol [42] and later adapted to Tilde [8], indicates which atoms can be added to a query, the maximal number of times the atom can be added (n 0) and the modes and types of the variables in it. A variable V in input mode, denoted with V , has to occur somewhere to the left in the query, whereas a variable in output mode, denoted with ....

....has proven to be particularly suitable for the design of upgrades to popular attribute value learning techniques. In that respect, Apriori Warmr is only one of the more recent additions to a sequence of similar upgrades [15] Explora [32] Claudien [16] CN2 [10] ICL [18] C4.5 [49] Tilde [8], and [33] C0.5 [14] 6.2 Clausal discovery Association rules A 1 : A k ) A k 1 : A n , as introduced in Section 3.1.2, can easily be confused with a clauses A 1 : A k A k 1 : A n : both are interpreted as if then rules with atoms A i . We first clarify the relation ....

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. In Proceedings of the Ninth Dutch Conference on Artificial Intelligence (NAIC97) , 1997.

.... and inconsistent theories that satisfy some other acceptance criteria (predictive accuracy, significance, compression) the predictive ILP setting can be extended to include learning of classification and prediction rules from imperfect data, as well as learning of logical decision trees [1]. In a broader sense, predictive ILP incorporates also first order regression [26] and constraint inductive logic programming [43] for which again different acceptance criteria apply. 2.2.2 Descriptive ILP Descriptive ILP is sometimes referred to as confirmatory induction, nonmonotonic ILP, ....

....and negative examples and background knowledge; in addition, Progol can also be used to learn from positive examples only. They use different acceptance criteria: compression, coverage accuracy and minimal description length, respectively. Induction of logical decision trees. The system Tilde [1] belongs to Top down induction of decision tree algorithms. It can be viewed as a firstorder upgrade of Quinlan s C4.5, employing logical queries in tree nodes which involves appropriate handling of variables. The main advantage of Tilde is its efficiency and capability of dealing with large ....

[Article contains additional citation context not shown here]

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Submitted to DAMI, Special Issue on Inductive Logic Programming, 1998.

....to be an option. In fact, this has been hinted at in [4] and in [14] In this paper we propose a method for generating the declarative language bias of top down ILP algorithms based on schemata. 2 Schemata as used in several ILP algorithms (e.g. FOCL [19] FOSSIL [12] SRT [14] and TILDE [2]) are a comparably useful and practical form of declarative bias for ILP, but their definition is still a hard and error prone task. The suggested solution amounts to a two level approach, where the user just has to declare the relationship between the meta level and the given database properly, ....

....be categorized as follows 3 : 1. Specification of complete clauses, meta level: Rule schemata [11] rule models [16] second order schemata [6] 2. Specification of refinement, meta level: Relational clich es [19, 18] 3. Specification of refinement, object level: FOSSIL [12] SRT [14] TILDE [2] The approach presented here belongs to the second group of methods, as it allows for the specification of refinements at the meta level. Next, we will compare it with probably the closest work in the literature, the approach based on relational clich es. Subsequently, we will compare it with two ....

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Technical Report Report CW 247, Katholieke Universiteit Leuven, Belgium, 1997.

....(39 chemicals) for testing. We also used available background knowledge to create additional representations. Experiments with monolithic machine learning systems As a first approach to solve the classification task, we trained and tested four different monolithic machine learning systems: Tilde (Blockeel Raedt 1997), Progol (Muggleton 1995) C4.5 (Quinlan 1993) and a naive bayesian classification (Smith 1988) system. system accuracy std. dev. C4.5 51.0 8.0 Progol 58.9 7.9 Naive Bayes 64.0 7.7 Tilde 71.8 7.2 Table 1: individual accuracies In order to evaluate the accuracy of the monolithic systems and ....

Blockeel, H., and Raedt, L. D. 1997. Top-down induction of logical decision trees. In Proceedings of the 9th Dutch Conference on Artificial Intelligence NAIC97.

....concept. However, recently the model based view of the ILP learning problem, which has originally been advocated for what has been called descriptional ILP [De Raedt and Dzeroski, 1994; Wrobel and Dzeroski, 1995] has also been adapted for classification learning [De Raedt and Van Laer, 1995; Blockeel and De Raedt, 1997] . In this framework, examples are interpretations, for which the learned theory has to be true [De Raedt, 1996] Many ILP learning problems can be formulated in both settings, which would yield different estimates, when the size of the example space is measured by merely counting the number of ....

Hendrik Blockeel and Luc De Raedt. Top-down induction of logical decision trees. Technical Report CW 247, Katholieke Universiteit Leuven, Department of Computer Science, Leuven, Belgium, January 1997.

....; V al n1 ] f166(X; V al 0166 ; V al n166 ] Table 1: The language definition for learning musk odor molecules. The results are summarized in Table 2, where we confront our results with the ones obtained by Dietterich et.al in [6] Also included in Table 2 is results from [3] where the relational decision tree learner TILDE was applied to the musk prediction task. TILDE was applied to the problem with different discretizations and language biases, the result we report here is the best result obtained in [3] Dietterich et.al hypothesised that a suitable representation ....

....et.al in [6] Also included in Table 2 is results from [3] where the relational decision tree learner TILDE was applied to the musk prediction task. TILDE was applied to the problem with different discretizations and language biases, the result we report here is the best result obtained in [3]. Dietterich et.al hypothesised that a suitable representation language in the musk domain consists of so called axis parallel hyperrectangles (APR) A hypothesis in this language simply specifies bounds for the molecule surface along the rays. In our parlance the representation language belongs ....

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Technical Report CW 247, Dept. of Computer Science, K.U.Leuven, January 1997.

.... At present, several inductive logic programming systems are available in the public domain and are free for academic purposes (cf. MLnet s web site at http: www.gmd.de ml archive) Well known inductive logic programming systems include Progol [16] Foil [17] RIBL [18] Claudien [19] Tilde [21] and Midos [20] Some of these systems (e.g. Progol, Midos, Tilde and Claudien) are being integrated and tested in commercial data mining systems such as Clementine (Integral Solutions Limited) and Kepler (Dialogis) Their implementation language is mostly C or Prolog. Due to difficulties with ....

Blockeel, H. and De Raedt, L.: Top-down induction of logical decision trees. Artificial Intelligence, Vol. 101, 1998.

No context found.

H. Blockeel, L. De Raedt, and J. Ramon. Top-down induction of logical decision trees. In Proc. 1998.

No context found.

H. Blockeel, L.D. Raedt. Top-down induction of logical decision trees. Artificial Intelligence, 1998.

No context found.

H. Blockeel and L. D. Raedt. Top-down induction of logical decision trees. Artificial Intelligence, 101(1-2), 1998.

No context found.

Hendrik Blockeel and Luc De Raedt. Top-down induction of logical decision trees. Artificial Intelligence, 101(1-2):285--297, 1998.

No context found.

H. Blockeel and L. De Raedt. Top-down induction of logical decision trees. Submitted to DAMI, Special Issue on Inductive Logic Programming, 1998.

No context found.

Hendrik Blockeel and Luc De Raedt. Top-down induction of logical decision trees. Artificial Intelligence, 101(1-2):285--297, 1998.

No context found.

H. Blockeel, L. De Raedt, and J. Ramon. Top-down induction of logical decision trees. In Proc. 1998.

No context found.

H. Blockeel and L. DeRaedt. Top-down induction of logical decision trees. Technical Report CW 247, Katholieke Universiteit Leuven, Belgium, 1997.

No context found.

H. Blockeel and L. DeRaedt. Top-down induction of logical decision trees. Technical Report Report CW 247, Katholieke Universiteit Leuven, Belgium, 1997.

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