Introducing possibilistic logic in ILP for dealing with exceptions

Attribute-value based representations, standard in today's data mining systems, have a limited expressiveness. Inductive Logic Programming provides an interesting alternative, particularly for learning from structured examples whose parts, each with its own attributes, are related to each other by means of first-order predicates. Several subsets of first-order logic (FOL) with different expressive power have been proposed in Inductive Logic Programming (ILP). The challenge lies in the fact that the more expressive the subset of FOL the learner works with, the more critical the dimensionality of the learning task. The Datalog language is expressive enough to represent realistic learning problems when data is given directly in a relational database, making it a suitable tool for data mining. Consequently, it is important to elaborate techniques that will dynamically decrease the dimensionality of learning tasks expressed in Datalog, just as Feature Subset Selection (FSS) techniques do it in attribute-value learning. The idea of re-using these techniques in ILP runs immediately into a problem as ILP examples have variable size and do not share the same set of literals. We propose here the first paradigm that brings Feature Subset Selection to the level of ILP, in languages at least as expressive as Datalog. The main idea is to first perform a change of representation, which approximates the original relational problem by a multi-instance problem. The representation obtained as the result is suitable for FSS techniques which we adapted from attribute-value learning by taking into account some of the characteristics of the data due to the change of representation. We present the simple FSS proposed for the task, the requisite change of representation, and the entire method combining those two algorithms. The method acts as a filter, preprocessing the relational data, prior to the model building, which outputs relational examples with empirically relevant literals. We discuss experiments in which the method was successfully applied to two real-world domains.

Model transformation by example is a novel trend in model-driven software engineering. The rationale behind this is to utilize existing knowledge represented by source and target models of previously developed systems; such as requirements analysis and software design models, respectively. Such knowledge can be utilized to derive transformation rules to be applied in future system developments. To achieve this goal, machine learning techniques can assist in discovering and formalizing desired transformation rules. Inductive Logic Programming (ILP) represents a highly applicable machine learning technique in this context. Given a set of examples and background knowledge encoded as a set of first-order logic descriptions, an ILP system attempts to derive rules describing different transformation steps in a purely declarative way. The induced rules follow the same logical description as the given examples and background knowledge. The objective of this work is to introduce initial setup of an ILP system that can be utilized to derive analysis-design transformation rules from a set of examples that represent pairs of analysis-design models.

"This book describes the aspects of mathematical logic related to computer sciences. The materials adopted in this book are intended to attend to both the peculiarities of logical systems and the requirements of computer science."

We now discuss the model theory of (classical) first-order logic together with partial interpretations. As in the propositional case, the syntax of partial firstorder logic is essentially the same as in the classical case. The main difference from classical first-order logic is the model theory, and the resulting formal axiomatics. The reader is referred e.g. to Smullyan [94] for a comprehensive discussion of classical first-order logic.

Learning first-order logic programs from relational facts yields intuitive insights into the data. Inductive logic programming (ILP) models are effective in learning first-order logic programs from observed relational data. Symbolic ILP models support rule learning in a data-ecient manner. However, symbolic ILP models are not robust to learn from noisy data. Neuro-symbolic ILP models utilize neural networks to learn logic programs in a differentiable manner which improves the robustness of ILP models. However, most neuro-symbolic methods need a strong language bias to learn logic programs, which reduces the usability and flexibility of ILP models and limits the logic program formats. In addition, most neuro-symbolic ILP methods cannot learn logic programs effectively from both small-size datasets and large-size datasets such as knowledge graphs. In the paper, we introduce a novel differentiable ILP model called differentiable first-order rule learner (DFORL), which is scalable to learn rules from both smaller and larger datasets. Besides, DFORL only needs the number of variables in the learned logic programs as input. Hence, DFORL is easy to use and does not need a strong language bias. We demonstrate that DFORL can perform well on several standard ILP datasets, knowledge graphs, and probabilistic relation facts and outperform several well-known differentiable ILP models. Experimental results indicate that DFORL is a precise, robust, scalable, and computationally cheap differentiable ILP model.

Recently, computer programs developed within the field of Inductive Logic Programming (ILP) have received some attention for their ability to construct restricted first-order logic solutions using problemspecific background knowledge. Prominent applications of such programs have been concerned with determining “structure-activity” relationships in the areas of molecular biology and chemistry. Typically the task here is to predict the “activity” of a compound, like toxicity, from its chemical structure. Research in the area shows that: (a) ILP programs have been restricted to qualitative predictions of activity (“high”, “low” etc.); (b) When appropriate attributes are available, ILP programs have not been able to better the performance of standard quantitative analysis techniques like linear regression. However ILP programs perform creditably when such attributes are unavailable; and (c) When both are applicable, ILP programs are usually slower than their propositional counterparts. This paper examines the use of ILP programs, not for obtaining theories complete for the sample, but as a method of “discovering” new attributes. These could then be used by methods like linear regression, thus allowing for quantitative predictions and the ability to use structural information as background knowledge. Using structure-activity tasks as a test-bed the utility of ILP programs in constructing new features was evaluated by examining the prediction of chemical activity using linear regression, with and without the aid of ILP learnt logical attributes. In three out of the five datasets examined the addition of ILP attributes produced statistically better results (P 10.01). In addition six important structural features that have escaped the attention of the expert chemists were discovered.KeywordsDihydrofolate ReductaseLogical AttributeInductive Logic ProgrammingBasic TemplateInductive Logic Programming SystemThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The paper investigates interpretations of propositional and first-order logic in which validity is defined in terms of partial indices; sometimes called possibilities but here understood as non-empty subsets of a set W of possible worlds. Truth at a set of worlds is understood to be truth at every world in the set. If all subsets of W are permitted the logic so determined is classical first-order predicate logic. Restricting allowable subsets and then imposing certain closure conditions provides a modelling for intuitionistic predicate logic. The same semantic interpretation rules are used in both logics for all the operators.

Classical first-order logic is in many ways central to work in mathematics, linguistics, computer science and artificial intelligence, so it is worthwhile to define it in full detail. We present soundness and completeness proofs of a sequent calculus for first-order logic, formalized in the interactive proof assistant Isabelle/HOL. Our formalization is based on work by Stefan Berghofer, which we have since updated to use Isabelle's declarative proof style Isar (Archive of Formal Proofs, Entry FOL-Fitting, August 2007 / July 2018). We represent variables with de Bruijn indices; this makes substitution under quantifiers less intuitive for a human reader. However, the nature of natural numbers yields an elegant solution when compared to implementations of substitution using variables represented by strings. The sequent calculus considered has the special property of an always empty antecedent and a list of formulas in the succedent. We obtain the proofs of soundness and completeness for the sequent calculus as a derived result of the inverse duality of its tableau counterpart. We strive to not only present the results of the proofs of soundness and completeness, but also to provide a deep dive into a programming-like approach to the formalization of first-order logic syntax, semantics and the sequent calculus. We use the formalization in a bachelor course on logic for computer science and discuss our experiences.

There exist valuable methods for theorem proving in non classical logics based on translation from these logics into first-order classical logic (abbreviated henceforth FOL). The key notion in these approaches istranslation from aSource Logic (henceforth abbreviated SL) to aTarget Logic (henceforth abbreviated TL). These methods are concerned with the problem offinding a proof in TL by translating a formula in SL, but they do not address the very important problem ofpresenting proofs in SL via a backward translation. We propose a framework for presenting proofs in SL based on a partial backward translation of proofs obtained in a familiar TL: Order-Sorted Predicate Logic. The proposed backward translation transfers some formulasF TL belonging to the proof in TL into formulasF SL , such that the formulasF SL either (a) belong to a corresponding deduction in SL (in the best case) or, (b) are semantically related in some precise way, to formulas in the corresponding deduction in SL (in the worst case). The formulasF TL andF SL can obviously be considered aslemmas of their respective proofs. Therefore the transfer of lemmas of TL gives at least a skeleton of the corresponding proof in SL. Since the formulas of a proof “keep trace” of the strategy used to obtain the proof, clearly the framework can also help in solving another fundamental and difficult problem:the transfer of strategies from classical to non classical logics. We show how to apply the proposed framework, at least to S5, S4(p), K, T, K4. Two conjectures are stated and we propose sufficient (and in general satisfactory) conditions in order to obtain formulas in the proof in SL. Two particular cases of the conjectures are proved to be theorems. Three examples are treated in full detail. The main lines of future research are given.

Concerning the three fundamental first-order fuzzy logics, the set of logically valid formulae is \(\varPi _2\)-complete for Łukasiewicz logic, \(\varPi _2\)-hard for Product logic, and \(\varSigma _1\)-complete for Godel logic, as with classical first-order logic. Among these fuzzy logics, only Godel logic is recursively axiomatisable. Hence, it was necessary to provide a hyperresolution-based proof method suitable for automated deduction, as one has done for classical logic. As another step, we can incorporate a countable set of intermediate truth constants of the form \(\bar{c}\), \(c\in (0,1)\), together with the equality, Open image in new window, strict order, Open image in new window, projection, Open image in new window, operators in Godel logic; and modify the hyperresolution calculus, inferring over so-called order clauses. We shall investigate the deduction problem of a formula from a countable theory in this expansion and the so-called canonical standard completeness, where the semantics of Godel logic is given by the standard \(\varvec{G}\)-algebra as well as truth constants are interpreted by ‘themselves’. The hyperresolution calculus is refutation sound and complete for a countable order clausal theory under a certain condition for the set of truth constants occurring in the theory. We get an affirmative solution to the open problem of recursive enumerability of unsatisfiable formulae in this expansion of Godel logic.

ConclusionRelational data mining based on inductive logic programming, first-order logic and probabilsitic estimates has several important advantages known from theoretical viewpoint Computational experiments presented in this chapter have shown these advantages practically for real financial data.Relational data mining methods and MMDR method, in particular, are able to discover useful regularities in financial time series for stock market prediction. In the time frames of the current study we obtained positive results using separately, SP500C and history of target itself for target forecast. The best of these regularities had shown about 75 % of correct forecasts on test data (1995–1996). The target variable was predicted using separately SP500 (close) and the target variable’s own history. Comparison of performance with other methods is presented in the next chapter.

A workmanlike, but nevertheless very effective combination of statistical and relational learning uses a statistical learner to construct models with features identified (quite often, separately) by a relational learner. This form of model-building has a long history in Inductive Logic Programming (ILP), with roots in the early 1990s with the LINUS system. Additional work has also been done in the field under the categories of propositionalisation and relational subgroup discovery, where a distinction has been made between elementary and non-elementary features, and statistical models have been constructed using one or the other kind of feature. More recently, constructing relational features has become an essential step in many model-building programs in the emerging area of Statistical Relational Learning (SRL). To date, not much work—theoretical or empirical—has been done on what kinds of relational features are sufficient to build good statistical models. On the face of it, the features that are needed are those that capture diverse and complex relational structure. This suggests that the feature-constructor should examine as rich a space as possible, in terms of relational descriptions. One example is the space of all possible features in first-order logic, given constraints of the problem being addressed. Practically, it may be intractable for a relational learner to search such a space effectively for features that may be potentially useful for a statistical learner. Additionally, the statistical learner may also be able to capture some kinds of complex structure by combining simpler features together. Based on these observations, we investigate empirically whether it is acceptable for a relational learner to examine a more restricted space of features than that actually necessary for the full statistical model. Specifically, we consider five sets of features, partially ordered by the subset relation, bounded on top by the set F d , the set of features corresponding to definite clauses subject to domain-specific restrictions; and bounded at the bottom by F e , the set of “elementary” features with substantial additional constraints. Our results suggest that: (a) For relational datasets used in the ILP literature, features from F d may not be required; and (b) Models obtained with a standard statistical learner with features from subsets of features are comparable to the best obtained to date.

Link mining problems are characterized by high complexity (since linked objects are not statistically independent) and uncertainty (since data is noisy and incomplete). Thus they necessitate a modeling language that is both probabilistic and relational. Markov logic provides this by attaching weights to formulas in first-order logic and viewing them as templates for features of Markov networks. Many link mining problems can be elegantly formulated and efficiently solved using Markov logic. Inference algorithms for Markov logic draw on ideas from satisfiability testing, Markov chain Monte Carlo, belief propagation, and resolution. Learning algorithms are based on convex optimization, pseudo-likelihood, and inductive logic programming. Markov logic has been used successfully in a wide variety of link mining applications and is the basis of the open-source Alchemy system.

This is a textbook based on an interdisciplinary (advanced undergraduate level) course that was originally taught by the three authors at Duke. It is intended for many different disciplines. Even though it covers mathematical and philosophical topics, it occasionally still feels like a textbook whose origins lie more in computer science than in the other disciplines. This isn’t even surprising, for an interdisciplinary focus on logic easily draws one towards computer science and artificial intelligence. In three parts, each written by one of the authors, a three-chapter overview is given of the topics of proof-theory, computability-theory and philosophical logic. Yet, rather than a juxtaposition of logical, mathematical, computational, and philosophical subjects, this textbook isn’t merely intended as an illustration of how logic is relevant for many different disciplines, but forms the basis for a different kind of introduction to logic. This reviewer surely applauds the intent to teach logic to philosophers more like we would teach it to computer scientists—with an emphasis on what it allows us to achieve, but also to teach it to computer scientists more like we would (or ought to) teach it to philosophers—with more attention to the diversity in logic than to the dominant background role of classical logic. While Part 1 is concerned with proof-theory, with chapters on classical propositional logic, first-order logic, and Prolog, it also double-tasks as an introduction to formal logic. Thus, chapter 1 reviews the language and semantics of propositional logic, and even goes on to prove two basic meta-theoretical results by

Inductive logic programming (ILP) realizes inductive machine learning in computational logic. However, the present ILP mostly handles classical clausal programs, especially Horn logic programs, and has limited applications to learning nonmonotonic logic programs . This article studies a method for realizing induction in nonmonotonic logic programs. We consider an extended logic program as a background theory, and introduce techniques for inducing new rules using answer sets of the program. The produced new rules explain positive/negative examples in the context of inductive logic programming. The proposed methods extend the present ILP techniques to a syntactically and semantically richer framework, and contribute to a theory of nonmonotonic ILP.