Quinlan, R. and Cameron-Jones, R. (1993). Foil: A midterm report. In European Conference on Machine Learning (ECML).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....set. More recently, a lazy propositionalization method has been proposed for the system PROPAL, which selectively propositionalizes the FOL training set by interleaving attribute value reformulation and algebraic resolution [3] The main representative of the second class of methods is FOIL 6. 0 [26], which automatically produces comparative literals of type V i k,V i # k, V i V j ,V i # V j , where V j are numerical variables already present in other non comparative literals and k is a numerical threshold. The semantics of the builtin relational predicates, as well as the heuristics for ....

J.R. Quinlan and R.M. Cameron-Jones, FOIL: A midterm report, in: Machine Learning: ECML-93, Lecture Notes in Artificial Intelligence, (Vol. 667), P.B. Brazdil, ed., Springer-Verlag, Berlin, 1993, pp. 3--20.

.... structure) and on the arti cial problem of Eastbound Trains proposed by Ryszard Michalski (prediction of trains directions based on their properties) For the two ILP domains, predictive accuracy is estimated by 10 and 5 fold cross validation, respectively, and results are compared to FOIL [20], Fors [9] and Progol. For the Eastbound Trains, the data is split into one training and test set partition, and the results are averaged over 8 iterations of the experiment. Predictive accuracy RIB is higher than or on par with the one of L. Pe na Castillo, unpublished, 2002 4 M. A. Krogel, ....

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proc. of the 6th European Conference on Machine Learning, 667: 3-20, 1993.

....relation instances (RDF triples) that is needed for the Semantic Web and the intelligent tools it promises. This paper explores the problem of learning information extraction rules that accurately derive ground facts characterising the content of natural language texts. We use the FOIL ILP learner[13], and therefore the problem becomes one of constructing the appropriate representation of the text, and of the background knowledge that is available. Naturally, this must be done automatically. The relations that are learned are those defined in a pre existing ontology of the domain. These ....

Quinlan, R.J. and Cameron-Jones, R.M. FOIL: A midterm report. Proc. European Conference on Machine Learning 1993.

....In this case, applying the CWA would yield a set S that contains all facts of the form v 1 (c 1 ; c 2 ; c 3 ; c 4 ) such that each c i appears in either S or B, but v 1 (c 1 ; c 2 ; c 3 ; c 4 ) 62 S . One off the shelf ILP system that supports use of the CWA is FOIL [Quinlan, 1990; Quinlan and Cameron Jones, 1993] FOIL allows one to specify that a set of positive examples is complete, and will automatically apply the CWA to any complete set of positive examples to obtain negative examples. FOIL also has a number of other features that are desirable for this problem: it is specialized to learn Datalog ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Pavel B. Brazdil, editor, Machine Learning: ECML-93, Vienna, Austria, 1993. Springer-Verlag. Lecture notes in Computer Science # 667.

....this area is to identify and analyze these special properties. For example, in many examples in which FOIL has learned recursive logic programs, it has made use of complete example sets datasets containing all examples of or below a certain size, rather than sets of randomly selected examples [Quinlan and Cameron Jones, 1993]. It is possible that complete datasets allow a more expressive class of programs to be learned than random datasets. Finally, and most importantly, this paper has established the boundaries of learnability for determinate recursive programs in the pac learnability model. In most plausible ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Pavel B. Brazdil, editor, Machine Learning: ECML-93, Vienna, Austria, 1993. Springer-Verlag. Lecture notes in Computer Science # 667.

.... 1996 ] with respect to their usual semantic interpretation of primitive concepts and roles, we show that reasoning in DL can be simulated by reasoning in horn logic with simple numeric constraints as formalized in [ Sebag and Rouveirol, 1996 ] and as present in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993 ] or Progol [ Muggleton, 1995 ] A simple invertible function encodes normalized concept descriptions into horn clauses using new predicates with an external semantics (as Borgida has show they are not expressible in horn logic) to represent the DL terms. The aim of this encoding, of course, is ....

.... i.e. should be # subsumed by #(C, X) for any concept description C, e.g. by any subclause containing the relation Formally this corresponds to reasoning and learning with simple numeric constraints as present in Constraint Logic Programs (CLP) In in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993] or Progol [ Muggleton, 1995 ] this is done with the help of the computed built in predicates and or #. For Foil or Progol a more suitable encoding is rrR (X, n atleast , matmost , Y ) as they can learn cmin and cmax in literals like cmin n atleast and matmost cmax . Sebag and ....

Quinlan, R. and R. M. Cameron-Jones: 1993, `FOIL: A Midterm Report'. In: P. Brazdil (ed.): Proceedings of the Sixth European Conference on Machine Leaning (ECML-93). Berlin, Heidelberg, pp. 3--20, Springer Verlag.

....can be obtained 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 ....

J.R. Quinlan and R.M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil (ed.), Proceedings of the 6th European Conference on Machine Learning, Vol. 667, pp. 3--20, Springer, 1993.

.... 1996 ] with respect to their usual semantic interpretation of primitive concepts and roles, we show that reasoning in DL can be simulated by reasoning in horn logic with simple numeric constraints as formalized in [ Sebag and Rouveirol, 1996 ] and as present in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993 ] or Progol [ Muggleton, 1995 ] A simple invertible function encodes normalized concept descriptions into horn clauses using new predicates with an external semantics (as Borgida has shown they are not expressible in horn logic) to represent the DL terms, i.e. from an ILP point of view, ....

..... Max 1 ] and [Min 2 . Max 2 ] is the interval [minimum(Min 1 , M in 2 ) maximum(Max 1 , Max 2 ) The handling of nothing, Formally this corresponds to reasoning and learning with simple numeric constraints as present in Constraint Logic Programs (CLP) In in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993 ] or Progol [ Muggleton, 1995 ] this is done with the help of the computed built in predicates and or #. For Foil or Progol a more suitable encoding is rrR (X, n atleast , matmost , Y ) as they can learn cmin and cmax in literals like cmin n atleast and matmost cmax . Sebag and ....

Quinlan, R. and R. M. Cameron-Jones: 1993, `FOIL: A Midterm Report'. In: P. Brazdil (ed.): Proceedings of the Sixth European Conference on Machine Leaning (ECML-93). Berlin, Heidelberg, pp. 3--20, Springer Verlag.

....it often generates a very large number of rules in association rule mining, and also it takes e#orts to select high quality rules from among them. In this paper, we propose a novel approach called CPAR (Classification based on Predictive Association Rules) CPAR inherits the basic idea of FOIL [9] in rule generation and integrates the features of associative classification in predictive rule analysis. In comparison with associative classification, CPAR has the following advantages: 1) CPAR generates a much smaller set of high quality predictive rules directly from the dataset; 2) to ....

....analysis. 3 Rule Generation The basic idea of CPAR is from FOIL, which is introduced in section 3.1. The rule generation algorithm of CPAR is developed step by step in sections 3.2 and 3.3. 3.1 FOIL: A Brief Introduction. FOIL (First Order Inductive Learner) proposed by Ross Quinlan in 1993 [9], is a greedy algorithm that learns rules to distinguish positive examples from negative ones. FOIL repeatedly searches for the current best rule and removes all the positive examples covered by the rule until all the positive examples in the data set are covered. The algorithm FOIL is presented ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Proc. 1993.

.... 1996 ] with respect to their usual semantic interpretation of primitive concepts and roles, we show that reasoning in DL can be simulated by reasoning in horn logic with simple numeric constraints as formalized in [ Sebag and Rouveirol, 1996 ] and as present in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993 ] or Progol [ Muggleton, 1995 ] A simple invertible function encodes normalized concept descriptions into horn clauses using new predicates with an external semantics (as Borgida has shown they are not expressible in horn logic) to represent the DL terms, i.e. from an ILP point of view, learning ....

.... and only if the encoding of C # I# subsumes the encoding of D (#(C) # I# #(D) and lcs(C, D) # 1 (lgg I# (#(C) #(D) Formally this corresponds to reasoning and learning with simple numeric constraints as present in Constraint Logic Programs (CLP) In in most ILP systems, e.g. Foil [ Quinlan and Cameron Jones, 1993] or Progol [ Muggleton, 1995 ] this is done with the help of the computed built in predicates and or #. For Foil or Progol a more suitable encoding is rrR (X, n atleast , matmost , Y ) as they can learn cmin and cmax in literals like cmin n atleast and matmost cmax . Sebag and ....

Quinlan, R. and R. M. Cameron-Jones: 1993, `FOIL: A Midterm Report'. In: P. Brazdil (ed.): Proceedings of the Sixth European Conference on Machine Leaning (ECML-93). Berlin, Heidelberg, pp. 3--20, Springer Verlag.

....the two main logical approaches: learning from entailment and learning from interpretations. The first one is considered the classical setting, also called explanatory ILP (Blockeel, 1998) and most of ILP systems use this setting (e.g. RDT (Kietz and Wrobel, 1992) Progol (Muggleton, 1995) FOIL (Quinlan, 1993a) SRT (Kramer, 1996) Fors (Karalic and Bratko, 1997) Learning from interpretations is also called descriptive ILP and more recently, ILP engines based on this setting have begun to appear (e.g. Claudien, ICL, and Tilde (De Raedt et al. 2001) An explanatory history of how learning from ....

....in Chapter 3 and future extensions in term of storage and speed efficiency that are currently being explored. Results of our experiments on the widely used Mutagenesis database indicate that MRDTL offers a promising alternative to existing algorithms such as Progol (Muggleton, 60 1995) FOIL (Quinlan, 1993a) and Tilde (Blockeel, 1998) Preliminary results of our experiments with the protein gene localization task (based on the data from the KDD Cup 2001 competition) indicate that MRDTL, if equipped with principled approaches to handling missing attribute values, can be an effective algorithm for ....

Quinlan, J. R. FOIL: A midterm report. In Proceedings of the 6 tn European Conference on Machine Learning, volume 667 of Lecture Notes in Artificial Intelligence, Springer-Verlag, 1993.

.... binary form (rule satisfied at least once) or as counts (the number of independent ways a training example can satisfy the first order rule) In this paper, we start to address the issues of combining ILP and regres sion by exploring some of the advantages and disadvantages of ILP using FOIL 4 [12, 13] and of logistic regression [9] separately and in combination, by look ing at the task of classifying scientific literature. We use the data from Re searchIndex, 5 an online digital library of computer science papers [1, 11] It con tains a rich set of relational data, including the text of ....

J.R. Quinlan and R.M. Cameron-Jones. FOIL: A midterm report. In Proceedings of the 6th European Conference on Machine Learning, pages 3-20, 1993.

.... in binary form (rule satis ed at least once) or as counts (the number of independent ways a training example can satisfy the rst order rule) In this paper, we start to address the issues of combining ILP and regression by exploring some of the advantages and disadvantages of ILP using FOIL [12, 13] and of logistic regression [9] separately, and in combination, by looking at the task of classifying scienti c literature. We use the data from ResearchIndex, an online digital library of computer science papers [1, 11] It contains a rich set of relational data, including the text of titles, ....

J.R. Quinlan and R.M. Cameron-Jones. FOIL: A midterm report. In Proceedings of the 6th European Conference on Machine Learning, pages 3-20, 1993.

....algorithms were used both for base level and meta level learning. For base level learning, they were applied to twenty datasets from the UCI Repository of Machine Learning Databases and Domain Theories [7] For meta level learning, the three propositional algorithms as well as two ILP systems FOIL [9] and TILDE [2] were applied to the results of base level learning, as described below. 3.1 Experimental Setting The measure of performance used in the experiments is the error rate of the classifier on the unseen examples. For each learning algorithm, the error rate for each of the twenty ....

Quinlan, J. R. and Cameron-Jones, R. M. (1993) FOIL: A midterm report. In Brazdil, P., editor: Proceedings of the 6th European Conference on Machine Learning, volume 667 of Lecture Notes in Artificial Intelligence, pages 3--20. SpringerVerlag.

....the diagram (the reason for this is explained in greater detail in Section 3) This procedure is approximate only because a nite number of alternatives are considered in Step 1. relev.tex; 12 03 2000; 14:45; p.3 In practice, executing Step 2 is not straightforward. It has been demonstrated (see [23]) that the presence of irrelevant background knowledge as could be the case here, for not all the background information need be used to construct all models can result in suboptimal or manifestly incorrect models. This is due to a combination of the restrictions built in to practical ILP systems ....

J. R. Quinlan. FOIL: a midterm report. In European Conference on Machine Learning. Springer-Verlag, Berlin, 1993.

....which statistical and other machine learning methods can aid the task. Suggestion 2: Encoding background knowledge The use of background knowledge to construct explanations is a distinctive feature of ILP. Results from experiments concerned with learning simple logic programs for list processing [27] suggest that the performance of an ILP system can be sensitive to the type and amount of background knowledge provided. Background knowledge that contains large amounts of information that is irrelevant to the problem being considered can, and typically does, hinder an ILP system in its search ....

J. R. Quinlan. FOIL: a midterm report. In European Conference on Machine Learning. Springer-Verlag, Berlin, 1993.

....of information known to be less relevant. 1 Introduction The use of background knowledge to construct explanations is a distinctive feature of Inductive Logic Programming (ILP: see [7] However, the results from experiments concerned with learning simple logic programs for list processing ([11]) suggest that the performance of an ILP system is sensitive to the type and amount of background knowledge provided. Background knowledge that contains large amounts of information that is irrelevant to the problem being considered can, and typically does, hinder an ILP system in its search for a ....

....from well established scientific and engineering knowledge. In this paper a classic chemistry task of finding structure activity relations has formed the basis for a systematic examination of the use of such knowledge by one particular ILP system. This examination was prompted by earlier work in [11] that showed that an ILP program could be adversely affected by large quantities of irrelevant information. This led to the question of how such programs behaved in more realistic situtations, where issues of relevance are less clear cut. Using an expert guided grading of background information, ....

J. R. Quinlan. FOIL: a midterm report. In European Conference on Machine Learning. Springer-Verlag, Berlin, 1993.

....implementing them listed below. 3.1 Decision Trees Decision trees are perhaps the most widely studied inductive learning models in the machine learning community. The literature abounds with papers proposing new models or variations of existing models and case studies using decision trees ([14, 21, 22, 25, 30, 34, 22 40, 43, 49, 50, 51, 53, 89, 93, 98, 99, 100, 101, 102, 104, 105, 106, 107, 109, 110, 111, 112, 113, 114, 118, 120, 123, 126, 129, 130, 131, 133, 134, 136]) For this case study, we use decision tree software from Quinlan and Buntine. Quinlan introduces decision trees and illustrates the use of his C4.5 software for decision trees (c4.5tree) and production rules derived therefrom (c4.5rule) in [105] Several decision tree algorithms (cart, id3, c4, ....

J.R. Quinlan and R.M. Cameron-Jones (1993). FOIL: A Midterm Report. ECML-93: Proceedings of the European Conference on Machine Learning Pavel Brazdil editor. Springer Verlag. 3-20.

....Publishers. Printed in the Netherlands. newpaper.tex; 6 07 2001; 20:11; p.1 the inclusion of human expertise, that has, in each case, decided the the background information to be used. This sorting of background knowledge into relevant and irrelevant can be extremely important: experiments in [20] suggest that background knowledge that contains large amounts of information that is irrelevant to the problem being considered can, and typically does, hinder an ILP system in its search for a model. Suciently skilled human experts may be capable of more than such a coarse sorting of background ....

J. R. Quinlan. FOIL: a midterm report. In European Conference on Machine Learning. Springer-Verlag, Berlin, 1993.

....of a graph, can be concisely represented using a first order representation. In this section, we consider the task of learning to classify pages using a learner that is able to induce first order rules. 5.2.1. Approach The learning algorithm that we use in this section is Quinlan s FOIL algorithm [52, 53]. FOIL is a greedy covering algorithm for learning function free Horn clauses. 4 FOIL induces each Horn clause by beginning with an empty tail and using a hill climbing search to add literals to the tail until the clause covers only (mostly) positive instances. The evaluation function used for ....

....hyperlink paths of unknown and variable size, we employ a learning method that uses a first order representation for its learned rules. Specifically, the algorithm we have developed for this task is based on the FOIL algorithm 94 M. Craven et al. Artificial Intelligence 118 (2000) 69 113 [52,53] which we used for page classification in Section 5.2. We discuss our algorithm in more detail below. The problem representation we use for this relation learning tasks consists of the following background relations: class(Page) For each class in the set of page classes considered in Section ....

J.R. Quinlan, R.M. Cameron-Jones, FOIL: A midterm report, in: Proc. European Conference on Machine Learning, Vienna, Austria, 1993, pp. 3--20.

....This research is partially supported by the Grant in Aid for Encouragement of Young Scientists No.08780346 from the Japanese Ministry of Education, Science, Sports and Culture. Our system FOIL I(First Order Inductive Learner from Incomplete samples) basically uses the framework of FOIL[QCJ93, Qui90]. FOIL I takes an extensional definition of a target relation and background relations described in Horn clauses. An extensional definition of a target relation consists of examples or tuples that satisfy the target relation and negative examples or tuples that do not satisfy it. The system ....

....difference length simultaneously. All partial clauses are kept in clauses without distinguishing their lengths. 2. More than one clauses can be found at the same time. FOIL prohibits to introduce a zero gain literal in principle, but some kind of literals are useful even if they are zero gain. [QCJ93] characterizes such kind of literals by using the idea of determinate literals originally proposed in [MF92] A 3 In default the m is ten. Initialization 1 theory : null program 2 remaining : all tuples belonging to target relation R 3 While remaining is not empty 4 clauses : f R(A; B; 1 ....

J.R. Quinlan and R.M. Cameron-Jones. FOIL: A midterm report. In P. Brazdil, editor, Proceedings of the 6th European Conference on Machine Learning, volume 667 of Lecture Notes in Artificial Intelligence, pages 3--20. Springer-Verlag, 1993.

....statistical approach. In this paper, we present a new approach to learning hypertext classi ers that combines a statistical text learning method with a relational rule learner. We present experiments that evaluate one particular instantiation of this general approach: a Foil based (Quinlan, 1990; Quinlan Cameron Jones, 1993) learner augmented with the ability to invent predicates using a Naive Bayes text classi er. Our experiments indicate that this approach is able to learn classi ers that are often more accurate than either purely statistical or purely relational alternatives. In previous research, the Web has ....

....to Hypertext Learning In this section we describe two approaches to learning in text domains. First we discuss the Naive Bayes algorithm, which is commonly used for text classi cation, and then we describe an approach that involves using a relational learning method, such as Foil(Quinlan, 1990; Quinlan Cameron Jones, 1993), for such tasks. These two algorithms are the constituents of the hybrid algorithm that we present in section 3. 2.1. Naive Bayes for Text Classi cation Most work in learning text classi ers involves representing documents as either sets of words or bags of words. Both of these are based on a ....

Quinlan, J. R., & Cameron-Jones, R. M. (1993). FOIL: A midterm report. In Proceedings of the Fifth European Conference on Machine Learning, pp. 3-20 Vienna, Austria. Springer-Verlag.

....programming many ILP systems demonstrate their abilities by showing that it is possible to generate simple Prolog programs at the level of the ones taught in a first logic programming course. As examples of some approaches concerned with automatic programming we can refer to the works of Quinlan [96,97], Bergadano et al. 5, 6] Flener [37,39] and Popelinsky et al. 94] Although currently the trend is to do inductive synthesis with logic programming languages such as Prolog, in the seventies and eighties the preferred language was LISP, which is closer to the functional paradigm. The works of ....

....Besides the examples, the system accepts type and input output mode declarations of the involved predicates. Dependency declarations between predicates are also given to the system. Background knowledge is defined intensionally. The systems GOLEM, by Muggleton and Feng [82] and FOIL, by Quinlan [96, 97, 98], were quite successful due to their relative efficiency and some practical problems to which these systems were applied. The GOLEM system induction engine is based in the lgg operator of Plotkin [92] which was already described here (Section 3.4.4) The system performs an incomplete bottom up ....

[Article contains additional citation context not shown here]

Quinlan, J. R. and Cameron-Jones, R. M. (1993): "FOIL: A Midterm Report". Proceedings of the European Conference on Machine Learning ECML-93. Ed. P. Brazdil. Springer-Verlag.

....of a graph, can be concisely represented using a first order representation. In this section, we consider the task of learning to classify pages using a learner that is able to induce first order rules. 5.2. 1 Approach The learning algorithm that we use in this section is Quinlan s Foil algorithm [52,53]. Foil is a greedy covering algorithm for learning function free Horn clauses 4 . Foil induces each Horn clause by beginning with an empty tail and using a hill climbing search to add literals to the tail until the clause covers only (mostly) positive instances. The evaluation function used for ....

....B. 6.1 Problem Representation Because this task involves discovering hyperlink paths of unknown and variable size, we employ a learning method that uses a first order representation for its learned rules. Specifically, the algorithm we have developed for this task is based on the Foil algorithm [52,53] which we used for page classification in Section 5.2. We discuss our algorithm in more detail below. The problem representation we use for this relation learning tasks consists of the following background relations: ffl class(Page) For each class in the set of page classes considered in ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Proceedings of the European Conference on Machine Learning, pages 3--20, Vienna, Austria, 1993.

....we derive a rule set from a decision tree by writing a rule for each path in the decision tree from the root to a leaf. 3.3 Learning relational rules We are searching for regularities in a relational knowledge base, and thus able to benefit from using a relational learner. Quinlan s FOIL system [15, 16] is a greedy covering algorithm for learning functionfree Horn clauses. FOIL induces each Horn clause by beginning with an empty tail and using a hill climbing search to add literals to the tail until the clause covers only (mostly) positive instances. The evaluation function used for the ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Proceedings of the European Conference on Machine Learning, pages 3--20, Vienna, Austria, 1993.

....is given, thus when dealing with numeric values, compelling to provide a bound on these values. If we consider again, the successor function and if we suppose that the domain is restricted to f0; 1; 2; 3g, we must either express that 3 has no successor, or introduce, as done in the system Foil [11], a new value expressing that the successor of 3 is out of range. Such requirements can lead to learn logic programs which take advantage of the closed world nature of the representation and which are incorrect when applied outside the nite domain. To deal with this problem, mostly generated ....

....the relation may not appear in every tuple. For instance, the two tuples true; true , true; false of a binary relation can be viewed as the generalized tuple X = true. 4 Our system 4. 1 The Learning Algorithm Our system uses a divide and conquer method, as the one used in the system Foil [11]: it learns a generalized relation covering some positive tuples, it removes the covered positive tuples, and the process is repeated until no more positive examples remains. Let POS and NEG respectively represent the positive and negative generalized tuples and let (RC ) represents a re nement ....

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Pavel B. Brazdil, editor, Proceedings of the European Conference on Machine Learning (ECML-93), volume 667 of LNAI, pages 3-20, Vienna, Austria, April 1993. Springer Verlag.

No context found.

Quinlan, J.R., and Cameron-Jones, R.M., FOIL: a midterm report, in: Proceedings European Conference on Machine Learning, Vienna (Springer-Verlag, Berlin, 1993) 3-20.

....order learning, continuous variables, attribute value representation, empirical evaluation. 1 1 Introduction First order learning systems like FOIL [Quinlan 1990, 1991] are commonly demonstrated on tasks such as learning recursive definitions of relations on lists of discrete constants, e.g. [Quinlan Cameron Jones, 1993] . Tasks of this kind seem far removed from the sort of attribute value learning tasks addressed by zeroth order systems such as C4.5 [Quinlan, 1993] particularly when continuous attributes are involved. However, FOCL [Silverstein Pazzani, 1991] introduces the notion of relational clich es, or ....

.... 1991] are commonly demonstrated on tasks such as learning recursive definitions of relations on lists of discrete constants, e.g. Quinlan Cameron Jones, 1993] Tasks of this kind seem far removed from the sort of attribute value learning tasks addressed by zeroth order systems such as C4.5 [Quinlan, 1993], particularly when continuous attributes are involved. However, FOCL [Silverstein Pazzani, 1991] introduces the notion of relational clich es, or strings of literals that form a pattern and so can be treated as a composite literal; these can include tests on continuous valued arguments. FOIL ....

[Article contains additional citation context not shown here]

Quinlan, J.R. and Cameron-Jones, R.M. (1993). FOIL: A midterm report. In Proceedings of the European Conference on Machine Learning, pages 3--20. Springer Verlag.

No context found.

Quinlan, R. and Cameron-Jones, R. (1993). Foil: A midterm report. In European Conference on Machine Learning (ECML).

No context found.

Quinlan, J.R., and Cameron-Jones, R.M., 1993. FOIL: a midterm report. In: P. Brazdil (Editor), Proc. European Conference on Machine Learning, Springer Verlag, pp. 3-20.

No context found.

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proc. of the 6th European Conference on Machine Learning, 667: 3-20, 1993.

No context found.

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proceedings of the 6th European Conference on Machine Learning, volume 667, pages 3-20. Springer-Verlag, 1993.

No context found.

Quinlan, J.R. (1993): FOIL: A Midterm Report. Proceedings of ECML-93, Springer-Verlag.

No context found.

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Proc. 1993.

No context found.

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In European Conf. Machine Learning, 1993.

No context found.

J. R. Quinlan and R. M. Cameron-Jones, "FOIL: A midterm report," in Machine Learning: ECML-93, 1993, pp. 3--20, http: //www.cse.unsw.edu.au/ # quinlan/q+cj.ecml93.ps. 87

No context found.

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proc. of the 6th European Conference on Machine Learning, 667: 3-20, 1993.

No context found.

Quinlan, J.R., Cameron-Jones, R.M.: FOIL: A midterm report, in Proceedings of the 6th European Conference on Machine Learning, LNAI 667, pp. 3-20, Springer-Verlag, 1993

No context found.

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In Proc. 1993.

No context found.

Quinlan, R.: FOIL: A midterm report. In: Proceedings of the 6th European Conference on Machine Learning, volume 667 of Lecture Notes in Artificial Intelligence, Springer-Verlag (1993)

No context found.

J. R. Quinlan and R. M. Cameron-Jones. FOIL: A midterm report. In P. B. Brazdil, editor, Machine Learning: ECML-93, Vienna, Austria, 1993. Springer-Verlag. Lecture notes in Computer Science # 667.

No context found.

J.R. Quinlan. FOIL: A midterm report. In P. Brazdil, editor, Proceedings of the 6th European Conference on Machine Learning, Lecture Notes in Arti cial Intelligence. Springer-Verlag, 1993.

No context found.

Quinlan, J. R., Cameron-Jones, R. M.: FOIL: A Midterm Report. Proc. of the European Conference on Machine Learning. (1993) 3--20

No context found.

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proceedings of the 6th European Conference on Machine Learning, volume 667, pages 3-20. Springer-Verlag, 1993.

No context found.

J. R. Quinlan. FOIL: a midterm report. In European Conference on Machine Learning. Springer-Verlag, Berlin, 1993.

No context found.

J.R. Quinlan and R. M. Cameron-Jones. FOIL: A Midterm Report. In P. Brazdil, editor, Proc. of the 6th European Conference on Machine Learning, volume 667, pages 3-20. Springer-Verlag, 1993.

No context found.

Quinlan, J.R., Cameron-Jones, R.M.: FOIL: A Midterm Report. In: Brazdil, P.B. #ed.#: Machine Learning: ECML-93, Lecture Notes in Arti#cial Intelligence, Vol. 667, Berlin: SpringerVerlag, Berlin Heidelberg New York #1993# 3-20.

No context found.

J.R. Quinlan. FOIL: A midterm report. In P. Brazdil, editor, Proceedings of the 6th European Conference on Machine Learning, Lecture Notes in Artificial Intelligence. Springer-Verlag, 1993.

No context found.

J.R. Quinlan, "FOIL: A Midterm Report", Proceedings of the Tenth International Conference on Machine Learning 1993, pp. 3-19

No context found.

J.R. Quinlan, "FOIL: A Midterm Report", Proceedings of the Tenth International Conference on Machine Learning 1993, pp. 3-19

First 50 documents  Next 50

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