Mostrando recursos 1 - 20 de 3.204

  1. Infinitary Logic and Inductive Definability Over Finite Structures

    Dawar, Anuj; Lindell, Steven; Weinstein, Scott
    The extensions of first-order logic with a least fixed point operators (FO + LFP) and with a partial fixed point operator (FO + PFP) are known to capture the complexity classes P and PSPACE respectively in the presence of an ordering relation over finite structures. Recently, Abiteboul and Vianu [AV91b] investigated the relation of these two logics in the absence of an ordering, using a mchine model of generic computation. In particular, they showed that the two languages have equivalent expressive power if and only if P = PSPACE. These languages can also be seen as fragments of an infinitary...

  2. Active Error Correction for learning kinship terms

    Morris, Gary
    Kinship Analysis requires learning definitions of all kinship terms used in a target culture from data that is gathered incrementally through interviews with informants. We exploit a collection of previously learned kinship definitions from different cultures (logical “domain theories”) to minimize the cost of learning the definitions in a new domain theory. We use Transfer Learning and Inductive Logic Programming (ILP) to learn kinship definitions. We propose a novel method for identifying potential errors in the data by comparing our data with definitions in previously learned models of similar cultures. We actively ask informants to confirm or correct these potential...

  3. FCSH

    David Botting; Waller Waller Guarini
    Abstract: I will show that there is a type of analogical reasoning that instantiates a pattern of reasoning in confirmation theory that is con-sidered at best paradoxical and at worst fatal to the entire syntactical approach to confirmation and ex-planation. However, I hope to elabo-rate conditions under which this is a sound (although not necessarily strong) method of reasoning. It does not, as its exponents claim, instanti-ate a pattern of reasoning distinct from deductive and inductive logic.

  4. Confirmation Theory and the Logic of Inductive Implication

    Ricardo S. Silvestre; Tarcísio H. C. Pequeno
    Abstract. The general purpose of this paper is to demonstrate through a well defined example how philosophy of science and Artificial Intelligence (AI) can benefit from each other by sharing some of their ideas, methods and techniques developed to tackle similar problems. The problem on which we will focus is the analysis of non-deductive inferences, which is performed in AI by the study of nonmonotonic commonsensical reasoning, and in philosophy of science by the so-called theory of inductive or evidential confirmation. After analyzing to what extent one of the most wide spread nonmonotonic formalisms – default logic – can be...

  5. M.O.: Learning Rules from Multisource Data for Cardiac Monitoring

    Élisa Fromont; René Quiniou; Marie-odile Cordier
    Abstract. This paper aims at formalizing the concept of learning rules from multisource data in a cardiac monitoring context. Our method has been implemented and evaluated on learning from data describing cardiac behaviors from different viewpoints, here electrocardiograms and arterial blood pressure measures. In order to cope with the dimensionality problems of multisource learning, we propose an Inductive Logic Programming method using a two-step strategy. Firstly, rules are learned independently from each sources. Secondly, the learned rules are used to bias a new learning process from the aggregated data. The results show that the the proposed method is much more...

  6. DOI 10.1007/s10994-006-8259-x Mathematical applications of inductive logic programming

    Mach Learn; Simon Colton; Stephen Muggleton
    highly successful. Such applications have led to breakthroughs in the domain of interest and have driven the development of ILP systems. The application of AI techniques to mathematical discovery tasks, however, has largely involved computer algebra systems and theorem provers rather than machine learning systems. We discuss here the application of the HR and Progol machine learning programs to discovery tasks in mathematics. While Progol is an established ILP system, HR has historically not been described as an ILP system. However, many applications of HR have required the production of first order hypotheses given data expressed in a Prolog-style manner,...

  7. Combining Macro-Operators with Control Knowledge

    Rocío García-durán; O Fernández; Daniel Borrajo
    Abstract. Inductive Logic Programming (ilp) methods have proven to succesfully acquire knowledge in very different learning paradigms, such as supervised and unsupervised learning or relational reinforcement learning. However, very little has been done on General Problem Solving (gps). One of the ilp-based approaches applied to gps is hamlet. This method is able to learn control rules (heuristics) for a non linear planner, prodigy4.0, which is integrated into the ipss system; control rules are used as an effective guide when building the planning search tree. Other learning approaches applied to planning generate macro-operators, building high-level blocks of actions, but increasing the...

  8. ILP with Noise and Fixed Example Size: A Bayesian Approach

    Current inductive logic programming systems are limited in their handling of noise, as they employ a greedy covering approach to constructing the hypothesis one clause at a time. This approach also causes difficulty in learning recursive predicates. Additionally, many current systems have an implicit expectation that the cardinality of the positive and negative examples reflect the "proportion " of the concept to the instance space. A framework for learning from noisy data and fixed example size is presented. A Bayesian heuristic for finding the most probable hypothesis in this general framework is derived. This approach evaluates a hypothesis as a...

  9. P.O.

    Eit Le

  10. Statistical Relational Learning and Probabilistic Inductive Logic Programming

    Stefano Bragaglia; Fabrizio Riguzzi
    Abstract. Logic Programs with Annotated Disjunctions (LPADs) are a promising language for Probabilistic Inductive Logic Programming. In order to develop efficient learning systems for LPADs, it is fundamental to have high-performing inference algorithms. The existing approaches take too long or fail for large problems. In this paper we adapt to LPAD the approaches for approximate inference that have been developed for ProbLog, namely k-best and Monte Carlo. k-Best finds a lower bound of the probability of a query by identifying the k most probable explanations while Monte Carlo estimates the prob-ability by smartly sampling the space of programs. The two...

  11. Helsinki 2009

    Ruurik Holm
    Constructive (intuitionist, anti-realist) semantics has thus far been lacking an adequate concept of truth in innity concerning factual (i.e., empirical, non-mathematical) sentences. One consequence of this problem is the difculty of incorporating inductive reasoning in constructive semantics. It is not possible to formulate a notion for probable truth in innity if there is no adequate notion of what truth in innity is. One needs a notion of a constructive possible world based on sensory expe-rience. Moreover, a constructive probability measure must be dened over these constructively possible empirical worlds. This study denes a particular kind of approach to the concept...

  12. Journal of Machine Learning Research xxx (2006) xxx-xxx Submitted 7/05; Revised 1/06; Published xxx/06 Kernels on Prolog Proof Trees: Statistical Learning in the ILP Setting ∗

    Andrea Passerini Passerini A○dsi·unifi·it; Paolo Frasconi P-f A○dsi·unifi·it; Luc De Raedt Deraedt A○informatik·uni-freiburg·de; Ute Schmid
    We develop kernels for measuring the similarity between relational instances using background knowledge expressed in first-order logic. The method allows us to bridge the gap between traditional inductive logic programming (ILP) representations and statistical approaches to supervised learning. Logic programs are first used to generate proofs of given visitor programs that use predicates declared in the available background knowledge. A kernel is then defined over pairs of proof trees. The method can be used for supervised learning tasks and is suitable for classification as well as regression. We report positive empirical results on Bongard-like and M-of-N problems that are difficult...

  13. Multi-Class Prediction Using Stochastic Logic Programs

    Jianzhong Chen; Lawrence Kelley; Stephen Muggleton; Michael Sternberg
    Abstract. In this paper, we present a probabilistic method of dealing with multiclass classification using Stochastic Logic Programs (SLPs), a Probabilistic Inductive Logic Programming (PILP) framework that integrates probability, logic representation and learning. Multi-class prediction attempts to classify an observed datum or example into its proper classification given that it has been tested to have multiple predictions. We apply an SLP parameter estimation algorithm to a previous study in the protein fold prediction area and a multi-class classification working example, in which logic programs have been learned by Inductive Logic Programming (ILP) and a large number of multiple predictions have...

  14. Solving selection problems using preference relation based on Bayesian learning

    Tomofumi Nakano; Nobuhiro Inuzuka
    Abstract. This paper defines a selection problem which selects an appropriate object from a set that is specified by parameters. We discuss inductive learning of selection problems and give a method combining inductive logic programming (ILP) and Bayesian learning. It induces a binary relation comparing likelihood of objects being selected. Our methods estimate probability of each choice by evaluating variance of an induced relation from an ideal binary relation. Bayesian learning combines a prior probability of objects and the estimated probability. By making several assumptions on probability estimation, we give several methods. The methods are applied to Part-of-Speech tagging. 1


    Toshiharu Takeuchi
    One of the hot research areas is knowledge discovery on structured documents like HTML and XML documents. In the case of XML documents, most popular approach to mining a knowledge is structural approach which find some kind of similar pattern(often tree structure or XPath) in interested XML documents. On the other hand, there is relational data mining approach such as ILP(Inductive Logic Programming). In this article relational approach for mining data from XML documents is proposed. To achieve relational data mining from XML documents, relations (or Background knowledge) about element or attribute in a XML documents is dynamically generated from...

  16. Toward Statistical Predicate Invention

    Stanley Kok; Pedro Domingos
    In the past few years, the statistical relational learning (SRL) community has recognized the importance of combining the strengths of statistical learning and relational learning (also known as inductive logic programming (ILP)), and developed several novel representations, as well as algorithms to learn their parameters and structure. However, the problem of statistical predicate invention (SPI) has so far received little attention in the community. SPI is the discovery of new concepts, properties and relations from data, expressed in terms of the observable ones, using statistical techniques to guide the process and explicitly representing the uncertainty in the discovered predicates. These...

  17. Cyclic proofs for first-order logic with inductive definitions

    James Brotherston
    Abstract. We consider a cyclic approach to inductive reasoning in the setting of first-order logic with inductive definitions. We present a proof system for this language in which proofs are represented as finite, locally sound derivation trees with a “repeat function ” identifying cyclic proof sections. Soundness is guaranteed by a well-foundedness condition formulated globally in terms of traces over the proof tree, following an idea due to Sprenger and Dam. However, in contrast to their work, our proof system does not require an extension of logical syntax by ordinal variables. A fundamental question in our setting is the strength...


    Bertil Rolf
    Whatever the nature of reasoning skills, such skills are rare [4], [2]. Thus, it would be desirable to develop support for them and to cultivate and strengthen them through proper education in reasoning. The background for my discussion is the development of support for reasoning skills that our research team has been conducting for some time ( Design of support or education for reasoning depends on concepts of reasoning skills. The essence of reasoning is to construct or evaluate relations of dependence. If D can be proved from A, B and C, there is a logical dependence between these items....

  19. 1 Annotation Concept Synthesis and Enrichment Analysis: a Logic- Based Approach to the Interpretation of High-Throughput Exper- iments

    Mikhail Jiline; Marcel Turcotte
    Motivation: Annotation Enrichment Analysis (AEA) is a widely used analytical approach to process data generated by high-throughput genomic and proteomic experiments such as gene expression mi-croarrays. The analysis uncovers and summarizes discriminating background information (e.g. GO annotations) for sets of genes identified by experiments (e.g. a set of differentially expressed genes, a cluster). The discovered information is utilized by human experts to find biological interpretations of the experiments. However, AEA isolates and tests for overrepresentation only individ-ual annotation terms or groups of similar terms and is limited in its ability to uncover complex phenomena involving relationship be-tween multiple annotation terms...

  20. Statistical Relational Learning and Probabilistic Inductive Logic Programming

    Stefano Bragaglia; Fabrizio Riguzzi
    Abstract. Logic Programs with Annotated Disjunctions (LPADs) are a promising language for Probabilistic Inductive Logic Programming. In order to develop efficient learning systems for LPADs, it is fundamental to have high-performing inference algorithms. The existing approaches take too long or fail for large problems. In this paper we adapt to LPAD the approaches for approximate inference that have been developed for ProbLog, namely k-Best and Monte Carlo. k-Best finds a lower bound of the probability of a query by identifying the k most probable explana-tions while Monte Carlo estimates the probability by smartly sampling the space of programs. The two...

Aviso de cookies: Usamos cookies propias y de terceros para mejorar nuestros servicios, para análisis estadístico y para mostrarle publicidad. Si continua navegando consideramos que acepta su uso en los términos establecidos en la Política de cookies.