S.Dzeroski, \Inductive Logic Programming and Knowledge Discovery in Databases", In Advances in KDD, MIT Press (1996) 117-152.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....is then more focussed or sowed towards a goal, and the knowledge that can be extracted is more expectable, and, consequently, less novel. The potential users of these tools are usually known as farmers [16] rather than miners or explorers . On the contrary, ILP approaches to data mining [7] [22] 6] allow the genera tion of richer concepts, since the expressiveness of logic programming is universal. For instance, the following rule can be induced by an ILP system without any transformation at all of the original database. good client(X) person(X) married(X,Y) good client(Y) ....

S. Dzeroski. Inductive Logic Programming and Knowledge Discovery in Databases. in Fayyad, U.; Pitatetsky-Shapiro, G, Smith, P. and Uthurusamy, R. (eds.) Ad- vances in Knowledge Discovery and Data Mining , MIT Press, Cambridge Mass., 1996.

....cannot neglect the implicit relations of spatial neighborhood (e.g. topological relations) that are defined by the explicit location and extension of spatial objects. As the interest in KDD is generally increasing, many recent applications of ILP methods and techniques to KDD have also emerged [3]. We claim that spatial data mining is a promising ILP application domain for two main reasons. First, ILP relies on the theory of computational logic which supplies representation and reasoning means appropriate for the spatial domain where relations among objects play a key role and are often ....

Dzeroski, S.: Inductive Logic Programming and Knowledge Discovery in Databases. In: Fayyad, U.M., Piatetsky-Shapiro, G., Smyth, P., Uthurusamy, R. (eds): Advances in Knowledge Discovery and Data Mining. AAAI Press/The MIT Press (1996) 117-152.

....being joined, in large domains this will generally not be an option. Many relational learners work by propositionalizing parts of the data on the fly (e.g. by adding attributes of related objects to the attributes of the objects of interest) and applying a propositional learner to the result [Dzeroski, 1996] . Doing this efficiently is a key but difficult problem, particularly when the relations involved do not all fit in main memory, and must be read from disk. We are currently addressing this using subsampling techniques in two ways [Hulten et al. 2003] The first is to minimize the number of ....

S. Dzeroski. Inductive logic programming and knowledge discovery in databases. In U. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pages 117--152. AAAI Press, Menlo Park, CA, 1996.

....in feature construction may stem from knowledge discovery in databases (KDD) 14] The given database representation has to be transformed into one which is accepted by the learning algorithm. Of course, for an ILP learning algorithm there exists a 1:1 mapping from a database table to a predicate[7]. However, this simple transformation most often is not one that eases learning. Since the arity of predicates cannot be changed through learning, a huge number of irrelevant database attributes is carried along. Therefore, the transformation from database tables into predicates should map parts ....

Saso Dzeroski. Inductive logic programming and knowledge discovery in databases. In Usama M. Fayyad, Gregory Piatetsky-Shapiro, Padhraic Smyth, and Ramasamy Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, chapter 1, pages 117--152. AAAI Press/The MIT Press, Menlo Park, California, 1996.

....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 was GOLEM [MF90] It can be classified as an empirical single predicate Inductive Logic Programming (ILP) learning system [Dze96] It is a batch non interactive system with noise handling capabilities that implements the relative 8 least general generalization principle that can be considered as careful generalization in the search space of possible concept descriptions. Rules generated by GOLEM can be processed in a ....

Saso Dzeroski. Inductive logic programming and knowledge discovery in databases. In U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pages 117-- 152. AAAI Press/MIT Press, Menlo Park, CA, 1996.

....Most of the currently available data mining algorithms are based on the so called attribute value (AV) approach which requires the reduction of a multi relational database to the single table format. The issue of mining from multiple relations has been extensively studied in the field of ILP [8]. Besides the elegant solution to multirelational data mining, the strenght of the ILP approach is the common background with deductive relational databases (DDB) which, e.g. can be fully exploited to implement the notion of inductive database [20] as pointed out by Flach [11] In recent times, a ....

....E F, if and only if there exist substitutions and such that E=F and F=E . We also say that E is an alphabetic variant of F. For instance, f(X) and f(Y) are alphabetic variants. learnability theory have suggested the use of prior knowledge and declarative bias to improve scalability [8, 25]. SPADA benefits from the available prior knowledge on the spatial domain and relies on a language bias specification to constrain the search for patterns. In particular, a refinement step consists of adding to the pattern to be refined one or more Datalog atoms in the language of patterns. ....

Dzeroski, S.: Inductive Logic Programming and Knowledge Discovery in Databases. In: Fayyad, U.M., Piatetsky-Shapiro, G., Smyth, P., Uthurusamy, R. (eds): Advances in Knowledge Discovery and Data Mining. AAAI Press/The MIT Press (1996) 117-152

....correlations between each cluster and each of a set of functional categories. Our analysis methodology is fundamentally different from the techniques discussed by Sherlock [Sherlock, 2000] We use the inductive logic programming (ILP) approach [Muggleton and Feng, 1990, Muggleton, 1999, Dzeroski, 1996] as an aid in data mining and formation of biological hypotheses. ILP is a technique that provides, in one integrated procedure, 1. A way to correlate output variables (gene expression) with input variables (functional categorizations, for instance) 2. A richer representational basis (allowing ....

Dzeroski, S. (1996). Inductive Logic Programming and Knowledge Discovery in Databases. In Fayyad, U., Piatetsky-Shapiro, G., Smyth, P., and Uthurusamy, R., editors, Advances in Knowledge Discovery and Data Mining, pages 117-152. AAAI Press/MIT Press.

....is most often used on relatively small data sets and the user has possibilities to participate during learning of clauses. Empirical ILP is automatic extraction of regularities from big and possibly imperferfect data sets without user interaction [Lav93] Overview of predicate learning in ILP [Dze96] Given . a set of positive (E ) and negative (E ) training examples ground facts of an unknown predicate p, a concept description language L , with syntactic restrictions on the definition of p, background knowledge B , defining predicates q i which may be used in the ....

Dzeroski S. Inductive Logic Programming and Knowledge Discovery in Databases. In Fayyad U.M., Piatetsky-Shapiro G., Smyth P. and Uthurusamy R., editor, Advances in Knowledge Discovery and Data Mining, pages 117--152. MIT Press, 1996. 2.2

....PYTHIA II uses a multimodal approach by integrating different learning methods to leverage their individual strengths. We have explored and implemented two such strategies: Case Based Reasoning (CBR) Joshi et al. 1996] and inductive logic programming (ILP) Bratko and Muggleton 1995; Dzeroski 1996; Muggleton and Raedt 1994] which we describe in this section. CBR systems obey a lazy learning paradigm in that learning consists solely of recording data from past experiments to help in future problemsolving sessions. This gain in simplicity of learning is offset by a more 232 . E. N. Houstis ....

....this definition is impractical, an approximate characterization, called the cover, is utilized which places greater emphasis on not representing the negative exemplars as opposed to representing the positive exemplars. Techniques such as relative least general generalization and inverse resolution [Dzeroski 1996] can then be applied to induce clausal definitions of the algorithm selection methodology. This forms the basis for building RS procedures using banks of selection rules. ILP is often prohibitively expensive, and the standard practice is to restrict the hypothesis space to a proper subset of ....

DZEROSKI, S. 1996. Inductive logic programming and knowledge discovery in databases. In U. FAYYAD,G.PIATETSKY-SHAPIRO,P.SMYTH, AND R. UTHURUSAMY Eds., Advances in Knowledge Discovery and Data Mining, pp. 117--152. AAAI Press/MIT Press.

....attempt to construct a predicate logic formula (such as patient(X,Y) so that all (most) positive examples can be logically derived from the background knowledge and no (few) negative examples can be logically derived. ILP has strong parallels in deductive database technology, as demonstrated in [Dzeroski, 1996]. Fig. 6 describes the results of correlating the mutagenicity of chemical compounds with their structure [Srinivasan and King, 1999b] As shown, the expressiveness of ILP [Bratko and Muggleton, 1995] makes it a highly desirable tool in structured domains where comprehension and interpretation of ....

Dzeroski, S. (1996). Inductive Logic Programming and Knowledge Discovery in Databases. In Fayyad, U., Piatetsky-Shapiro, G., Smyth, P., and Uthurusamy, R., editors, Advances in Knowledge Discovery and Data Mining, pages 117--152. AAAI/MIT Press.

....neither approach is very attractive, especially from an efficiency point of view, and a proper way of dealing with multiple relations is necessary. The idea of mining from multiple tables is not a new one. It is being studied extensively in the field of Inductive Logic Programming (ILP) [6, 15]. However, these approaches are mostly based on data stored as Prolog programmes, and little attention has been given to data stored in relational database and to how knowledge of the data model can help to guide the search process [1, 16, 21] Nor has a lot of attention been given to efficiency ....

Dzeroski, S. Inductive Logic Programming and Knowledge Discovery in Databases, Advances in Knowledge Discovery and Data Mining, AAAI Press, 1996

....we have integrated mostly systems that generate comprehensible knowledge in the form of logic rules, if then else rules or even decision trees. The first system we integrated and for which we will present some results later on, was GOLEM [Muggleton and Feng 1990] which has been classified 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 the relative least general generalization principle that can be considered as careful generalization in the search space of ....

Dzeroski, S. 1996. Inductive Logic Programming and Knowledge Discovery in Databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy (Eds.), Advances in Knowledge Discovery and Data Mining, pp. 117--152. AAAI Press/MIT Press.

....area for machine learning is knowledge discovery from large databases with rich data formats, which might contain for example satellite images, audio recordings, movie les, etc. While Dzeroski has shown how ILP applies very naturally to knowledge discovery from ordinary relational databases [6], advances are needed to deal with multimedia databases. ILP has advantages over other machine learning techniques for all of the preceding application areas. Nevertheless, these and other potential applications also highlight the following shortcomings of present ILP technology. Fig. 2. ACE ....

S. Dzeroski. Inductive logic programming and knowledge discovery in databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining. 1996.

....PYTHIA II uses a multi modal approach by integrating di#erent learning methods to leverage their individual strengths. We have explored and implemented two such strategies: Case Based Reasoning (CBR) Joshi et al. 1996] and inductive logic programming (ILP) Bratko and Muggleton 1995; Dzeroski 1996; Muggleton and Raedt 1994] which we describe in this section. CBR systems obey a lazy learning paradigm in that learning consists solely of recording data from past experiments to help in future problem solving sessions. This gain in simplicity of learning is o#set by a more complicated process ....

....this definition is impractical, an approximate characterization, called the cover, is utilized which places greater emphasis on not representing the negative exemplars as opposed to representing the positive exemplars. Techniques such as relative least general generalization and inverse resolution [Dzeroski 1996] can then be applied to induce clausal definitions of the algorithm selection methodology. This forms the basis for building RS procedures using banks of selection rules. ILP is often prohibitively expensive and the standard practice is to restrict the hypothesis space to a proper subset of first ....

Dzeroski, S. 1996. Inductive logic programming and knowledge discovery in databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy (Eds.), Advances in Knowledge Discovery and Data Mining, pp. 117--152. AAAI Press/MIT Press.

....strengths. The PYTHIA II system is a general framework enabling the integration of a range of reasoning and learning techniques. We have explored and implemented two such strategies: Case Based Reasoning (CBR) Joshi et al. 1996] and inductive logic programming (ILP) Bratko and Muggleton 1995; Dzeroski 1996; Muggleton and Raedt 1994] In the remainder of this section, we describe the CBR and ILP approaches and explain their use. Such learning and reasoning systems can typically be characterized as either lazy learning or eager learning paradigms. CBR systems obey a lazy learning paradigm in ....

....this definition is impractical, an approximate characterization, called the cover, is utilized which places greater emphasis on not representing the negative exemplars as opposed to representing the positive exemplars. Techniques such as relative least general generalization and inverse resolution [Dzeroski 1996] can then be applied to induce clausal definitions of the algorithm selection methodology. This forms the basis for building RS procedures using banks of selection rules. ILP is often prohibitively expensive and the standard practice is to restrict the hypothesis space to a proper subset of first ....

[Article contains additional citation context not shown here]

Dzeroski, S. 1996. Inductive logic programming and knowledge discovery in databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy (Eds.), Advances in Knowledge Discovery and Data Mining, pp. 117--152. AAAI Press/MIT Press.

....the parents, and that we will be considering different sets of parents in order to come up with good indicators for classifying parents as car owners or not. The idea of mining from multiple tables is not a new one. It is being studied extensively in the field of Inductive Logic Programming (ILP) [6, 11]. However, these approaches are mostly based on data stored as Prolog programs, and little attention has been given to data stored in relational database and to how knowledge of the data model can help to guide the search process [1, 12, 19] Nor has a lot of attention been given to efficiency and ....

Dzeroski, S. Inductive Logic Programming and Knowledge Discovery in Databases, Advances in Knowledge Discovery and Data Mining, AAAI Press, 1996

....strengths. The PYTHIA#II system is a general framework enabling the integration of a range of reasoning and learning techniques. We have explored and implemented two such strategies# Case#Based Reasoning #CBR# #Joshi et al. 1996# and inductive logic programming #ILP# #Bratko and Muggleton 1995# Dzeroski 1996# Muggleton and Raedt 1994#. In the remainder of this section# we describe the CBR and ILP approaches and explain their use. Such learning and reasoning systems can typically be characterized as either #lazy learning# or #eager learning# paradigms. CBR systems obey a #lazy#learning# paradigm in ....

....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 careful generalization in the search space ....

Dzeroski# S. 1996. Inductive logic programming and knowledge discovery in databases. In U. Fayyad# G. Piatetsky#Shapiro# P. Smyth# and R. Uthurusamy #Eds.## Advances in Knowledge Discovery and Data Mining# pp. 117#152. AAAI Press#MIT Press.

....techniques is planned for the near future. 6 Discussion and Related work We here touch upon some related work. We restrict ourselves to research not explicitly addressed elsewhere in the paper. For an overview of ILP work in the context of knowledge discovery in databases, we refer to [23]. 6.1 Logical paradigm: learning from interpretations The definition of frequent query discovery and the (relatively) efficient candidate evaluation in Warmr is rooted in the learning from interpretations paradigm, introduced by De Raedt and Dzeroski [17] and related to other inductive logic ....

S. Dzeroski. Inductive logic programming and knowledge discovery in databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pages 118--152. The MIT Press, 1996.

....databases. Furthermore, it may be possible to learn context in one stage, and then apply relational learning approaches in a later stage. One such possibility is to make use of inductive learners that learn concept descriptions in first order logic, such as FOIL (Quinlan 1990) or ILP methods (Dzeroski 1996). Given the appropriate context information, it is possible that more expressive methods could learn general relational rules such as the following, which indicates fraud when a user calls from an abnormal location. CALL ORIGIN(X,ORIG) NORMAL CALL LOCS(USER,LOCS) ORIG = 2 LOCS = FRAUD ....

Dzeroski, S. (1996). Inductive logic programming and knowledge discovery in databases. In Advances in Knowledge Discovery and Data Mining, Menlo Park, CA, pp. 117--152. AAAI Press.

....algorithms is so far restricted to databases that consist of a single relation composed of a set of binary attributes. We describe how these restrictions can be overcome through the combination of the available algorithms with standard techniques from the field of inductive logic programming [D zeroski, 1996; De Raedt, 1996a; Lavrac and Dzeroski, 1994] In Section 2, we first formalize the concept of association rules over multiple relations and the task of mining them in a deductive relational database. In Section 3, we present the algorithm AprioriRel, which extends Apriori [Agrawal et al. 1996] ....

S. Dzeroski. Inductive logic programming and knowledge discovery in databases. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pages 118--152. The MIT Press, 1996.

....knowledge to an attribute value representation (propositional logic) possibly with the extension of probabilistic annotations or certainty factors. This is a severe restriction, since it excludes relational or structural knowledge. Saso Dzeroski has given a nice example to illustrate this [3]: Figure 1 shows data in two tables of a database. If restricted to propositional knowledge, an algorithm could discover the following rules in the data: income(P erson) 100000 customer(P erson) yes sex(P erson) f age(P erson) 32 customer(P erson) yes Rules like the following cannot ....

Saso Dzeroski. Inductive logic programming and knowledge discovery in databases. In Usama M. Fayyad et al., editors, see 4., pages 117--152.

No context found.

S.Dzeroski, \Inductive Logic Programming and Knowledge Discovery in Databases", In Advances in KDD, MIT Press (1996) 117-152.

No context found.

S. Dzeroski, "Inductive logic programming and knowledge discovery in databases," in Advances in Knowledge Discovery and Data Mining, U. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, Eds. Cambridge, MA: AAAI Press/MIT Press, 1996.

No context found.

S. Dzeroski. Inductive logic programming and knowledge discovery in databases. In U. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, chapter 5, pages 117--152. MIT Press, 1996.

First 50 documents

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