We extend the notion of anti-unification to cover equational theories and present a method based on regular tree grammars to compute a finite representation of E-generalization sets. We present a framework to combine ...
详细信息
We extend the notion of anti-unification to cover equational theories and present a method based on regular tree grammars to compute a finite representation of E-generalization sets. We present a framework to combine inductive logic programming and E-generalization that includes an extension of Plotkin's lgg theorem to the equational case. We demonstrate the potential power of E-generalization by three example applications: computation of suggestions for auxiliary lemmas in equational inductive proofs, computation of construction laws for given term sequences, and learning of screen editor command sequences. (c) 2005 Elsevier B.V. All rights reserved.
With the advent of the Semantic Web, description logics have become one of the most prominent paradigms for knowledge representation and reasoning. Progress in research and applications, however, is constrained by the...
详细信息
With the advent of the Semantic Web, description logics have become one of the most prominent paradigms for knowledge representation and reasoning. Progress in research and applications, however, is constrained by the lack of well-structured knowledge bases consisting of a sophisticated schema and instance data adhering to this schema. It is paramount that suitable automated methods for their acquisition, maintenance, and evolution will be developed. In this paper, we provide a learning algorithm based on refinement operators for the description logic ALCQ including support for concrete roles. We develop the algorithm from thorough theoretical foundations by identifying possible abstract property combinations which refinement operators for description logics can have. Using these investigations as a basis, we derive a practically useful complete and proper refinement operator. The operator is then cast into a learning algorithm and evaluated using our implementation DL-Learner. The results of the evaluation show that our approach is superior to other learning approaches on description logics, and is competitive with established ILP systems.
Relational reinforcement learning is presented, a learning technique that combines reinforcement learning with relational learning or inductive logic programming. Due to the use of a more expressive representation lan...
详细信息
Relational reinforcement learning is presented, a learning technique that combines reinforcement learning with relational learning or inductive logic programming. Due to the use of a more expressive representation language to represent states, actions and Q-functions, relational reinforcement learning can be potentially applied to a new range of learning tasks. One such task that we investigate is planning in the blocks world, where it is assumed that the effects of the actions are unknown to the agent and the agent has to learn a policy. Within this simple domain we show that relational reinforcement learning solves some existing problems with reinforcement learning. In particular, relational reinforcement learning allows us to employ structural representations, to abstract from specific goals pursued and to exploit the results of previous learning phases when addressing new (more complex) situations.
The paper introduces LOGAN-H-a system for learning first-order function-free Horn expressions from interpretations. The system is based on an algorithm that learns by asking questions and that was proved correct in pr...
详细信息
The paper introduces LOGAN-H-a system for learning first-order function-free Horn expressions from interpretations. The system is based on an algorithm that learns by asking questions and that was proved correct in previous work. The current paper shows how the algorithm can be implemented in a practical system, and introduces a new algorithm based on it that avoids interaction and learns from examples only. The LOGAN-H system implements these algorithms and adds several facilities and optimizations that allow efficient applications in a wide range of problems. As one of the important ingredients, the system includes several fast procedures for solving the subsumption problem, an NP-complete problem that needs to be solved many times during the learning process. We describe qualitative and quantitative experiments in several domains. The experiments demonstrate that the system can deal with varied problems, large amounts of data, and that it achieves good classification accuracy.
inductive logic programming (ILP) and Relational Data Mining (RDM) address the task of inducing models or patterns from multi-relational data. One of the established approaches to RDM is propositionalization, characte...
详细信息
inductive logic programming (ILP) and Relational Data Mining (RDM) address the task of inducing models or patterns from multi-relational data. One of the established approaches to RDM is propositionalization, characterized by transforming a relational database into a single-table representation. This paper presents a propositionalization technique called wordification which can be seen as a transformation of a relational database into a corpus of text documents. Wordification constructs simple, easy to understand features, acting as words in the transformed Bag-Of-Words representation. This paper presents the wordification methodology, together with an experimental comparison of several propositionalization approaches on seven relational datasets. The main advantages of the approach are: simple implementation, accuracy comparable to competitive methods, and greater scalability, as it performs several times faster on all experimental databases. Furthermore, the wordification methodology and the evaluation procedure are implemented as executable workflows in the web-based data mining platform ClowdFlows. The implemented workflows include also several other ILP and RDM algorithms, as well as the utility components that were added to the platform to enable access to these techniques to a wider research audience. (C) 2015 Elsevier Ltd. All rights reserved.
A rule-based program will return a set of answers to each query. An impure program, which includes the Prolog cut "!" and "not(.)" operators, can return different answers if its rules are re-ordere...
详细信息
A rule-based program will return a set of answers to each query. An impure program, which includes the Prolog cut "!" and "not(.)" operators, can return different answers if its rules are re-ordered. There are also many reasoning systems that return only the first answer found for each query;these first answers, too, depend on the rule order, even in pure rule-based systems. A theory revision algorithm, seeking a revised rule-base whose expected accuracy, over the distribution of queries, is optimal, should therefore consider modifying the order of the rules. This paper first shows that a polynomial number of training "labeled queries" (each a query paired with its correct answer) provides the distribution information necessary to identify the optimal ordering. It then proves, however, that the task of determining which ordering is optimal, once given this distributional information, is intractable even in trivial situations;e.g., even if each query is an atomic literal, we are seeking only a "perfect" theory, and the rule base is propositional. We also prove that this task is not even approximable: Unless P = NP, no polynomial time algorithm can produce an ordering of an n-rule theory whose accuracy is within n(gamma) of optimal, for some gamma > 0. We next prove similar hardness and non-approximatability, results for the related tasks of determining, in these impure contexts, (1) the optimal ordering of the antecedents;(2) the optimal set of new rules to add and (3) the optimal set of existing rules to delete. (C) 1999 Elsevier Science Inc. All rights reserved.
A magic value in a program is a constant symbol that is essential for the execution of the program but has no clear explanation for its choice. Learning programs with magic values is difficult for existing program syn...
详细信息
A magic value in a program is a constant symbol that is essential for the execution of the program but has no clear explanation for its choice. Learning programs with magic values is difficult for existing program synthesis approaches. To overcome this limitation, we introduce an inductive logic programming approach to efficiently learn programs with magic values. Our experiments on diverse domains, including program synthesis, drug design, and game playing, show that our approach can (1) outperform existing approaches in terms of predictive accuracies and learning times, (2) learn magic values from infinite domains, such as the value of pi, and (3) scale to domains with millions of constant symbols.
Two representation changes are presented: the first one, called flattening, transforms a first-order logic program with function symbols into an equivalent logic program without function symbols;the second one, called...
详细信息
Two representation changes are presented: the first one, called flattening, transforms a first-order logic program with function symbols into an equivalent logic program without function symbols;the second one, called saturation, completes an example description with relevant information with respect to both the example and available background knowledge. The properties of these two representation changes are analyzed as well as their influence on a generalization algorithm that takes a single example as input.
Recently there has been growing interest both to extend ILP to description logics and to apply it to knowledge discovery in databases. In this paper we present a novel approach to association rule mining which deals w...
详细信息
Recently there has been growing interest both to extend ILP to description logics and to apply it to knowledge discovery in databases. In this paper we present a novel approach to association rule mining which deals with multiple levels of description granularity. It relies on the hybrid language AL-log which allows a unified treatment of both the relational and structural features of data. A generality order and a downward refinement operator for AL-log pattern spaces is defined on the basis of query subsumption. This framework has been implemented in SPADA, an ILP system for mining multi-level association rules from spatial data. As an illustrative example, we report experimental results obtained by running the new version of SPADA on geo-referenced census data of Manchester Stockport.
Traffic incidents inevitably cause traffic delay and deteriorate road safety conditions. Incidents are increasing alongside the fast economic growth. Due to the rampant growth of traffic incidents, developing efficien...
详细信息
Traffic incidents inevitably cause traffic delay and deteriorate road safety conditions. Incidents are increasing alongside the fast economic growth. Due to the rampant growth of traffic incidents, developing efficient and effective automated incident detection (AID) techniques has prompted a growing worldwide interest. In this paper, the great efforts on developing a new approach to this problem based on nFOIL, a novel inductive logic programming (ILP), are done. By way of illustration, a simulated traffic data generated from Ayer Rajah Expressway (AYE) highway in Singapore and a real traffic data collected in I-880 freeway in California are used to assess the detection performance of this approach, and performance metrics includes detection rate, false alarm rate, mean time to detection, classification rate and the area under Receiver Operating Characteristic (ROC) curve (AUC). For comparison, we conducted the experiments on neural networks and support vector machine. The experimental results showed that nFOIL is sensitive to the skewed distribution of positive and negative examples in the dataset, and we make use of two different techniques, resampling and ensemble learning, to cope with highly skewed data in the context of ILP classification problems and investigated the effect of them typicality on the performance of AID model. It is concluded that ILP based AID approach are feasible, and have a favorable performance compared to neural networks and support vector machines. (C) 2011 Elsevier Ltd. All rights reserved.
暂无评论