Bart Demoen

On the implementation of the probabilistic logic programming language problog

The past few years have seen a surge of interest in the field of probabilistic logic learning and statistical relational learning. In this endeavor, many probabilistic logics have been developed. ProbLog is a recent probabilistic extension of Prolog motivated by the mining of large biological networks. In ProbLog, facts can be labeled with probabilities. These facts are treated as mutually independent random variables that indicate whether these facts belong to a randomly sampled program. Different kinds of queries can be posed to ProbLog programs. We introduce algorithms that allow the efficient execution of these queries, discuss their implementation on top of the YAP-Prolog system, and evaluate their performance in the context of large networks of biological entities.

Improving the efficiency of inductive logic programming through the use of query packs

Inductive logic programming, or relational learning, is a powerful paradigm for machine learning or data mining. However, in order for ILP to become practically useful, the efficiency of ILP systems must improve substantially. To this end, the notion of a query pack is introduced: it structures sets of similar queries. Furthermore, a mechanism is described for executing such query packs. A complexity analysis shows that considerable efficiency improvements can be achieved through the use of this query pack execution mechanism. This claim is supported by empirical results obtained by incorporating support for query pack execution in two existing learning systems.

Scaling Up Inductive Logic Programming by Learning from Interpretations

When comparing inductive logic programming (ILP) and attribute-value learning techniques, there is a trade-off between expressive power and efficiency. Inductive logic programming techniques are typically more expressive but also less efficient. Therefore, the data sets handled by current inductive logic programming systems are small according to general standards within the data mining community. The main source of inefficiency lies in the assumption that several examples may be related to each other, so they cannot be handled independently.Within the learning from interpretations framework for inductive logic programming this assumption is unnecessary, which allows to scale up existing ILP algorithms. In this paper we explain this learning setting in the context of relational databases. We relate the setting to propositional data mining and to the classical ILP setting, and show that learning from interpretations corresponds to learning from multiple relations and thus extends the expressiveness of propositional learning, while maintaining its efficiency to a large extent (which is not the case in the classical ILP setting).As a case study, we present two alternative implementations of the ILP system TILDE (Top-down Induction of Logical DEcision trees): TILDEclassic, which loads all data in main memory, and TILDELDS, which loads the examples one by one. We experimentally compare the implementations, showing TILDELDS can handle large data sets (in the order of 100,000 examples or 100 MB) and indeed scales up linearly in the number of examples.

The computational power and complexity of constraint handling rules

Constraint Handling Rules (CHR) is a high-level rule-based programming language which is increasingly used for general-purpose programming. We introduce the CHR machine, a model of computation based on the operational semantics of CHR. Its computational power and time complexity properties are compared to those of the well-understood Turing machine and Random Access Memory machine. This allows us to prove the interesting result that every algorithm can be implemented in CHR with the best known time and space complexity. We also investigate the practical relevance of this result and the constant factors involved. Finally we expand the scope of the discussion to other (declarative) programming languages.

So Many WAM Variations, So Little Time

The WAM allows within its framework many variations e.g. regarding the term representation, the instruction set and the memory organization. Consequently several Prolog systems have implemented successful variants of the WAM. While these variants are effective within their own context, it is difficult to assess the merit of their particular variation. In this work, four term representations that were used by at least one successful system are compared empirically within dProlog, one basic implementation which keeps all other things equal. We also report on different implementation choices in the dProlog emulator itself. dProlog is reasonably efficient, so it makes sense to use it for these experiments.

An Overview of HAL

Experience using constraint programming to solve real-life problems has shown that finding an efficient solution to the problem often requires experimentation with different constraint solvers or even building a problem-specific constraint solver. HAL is a new constraint logic programming language expressly designed to facilitate this process. It provides a well-defined solver interface, mutable global variables for implementing a constraint store, and dynamic scheduling for combining, extending and writing new constraint solvers. Equally importantly, HAL supports semi-optional type, mode and determinism declarations. These allow natural constraint specification by means of type overloading, better compile-time error checking and generation of more efficient run-time code.

On the Efficient Execution of ProbLog Programs

The past few years have seen a surge of interest in the field of probabilistic logic learning or statistical relational learning. In this endeavor, many probabilistic logics have been developed. ProbLog is a recent probabilistic extension of Prolog motivated by the mining of large biological networks. In ProbLog, facts can be labeled with mutually independent probabilities that they belong to a randomly sampled program. Different kinds of queries can be posed to ProbLog programs. We introduce algorithms that allow the efficient execution of these queries, discuss their implementation on top of the YAP-Prolog system, and evaluate their performance in the context of large networks of biological entities.

Analyzing logic programs using “prop”-ositional logic programs and a magic wand

Abstract This paper illustrates the role of a class of “prop”-ositional logic programs in the analysis of complex properties of logic programs. Analyses are performed by abstracting Prolog programs to corresponding “prop”-ositional logic programs which approximate the original programs and have finite meanings. We focus on a groundness analysis which is equivalent to that obtained by abstract interpretation using the domain Prop . The main contribution is in the ease in which a highly efficient implementation of the analysis is obtained. The implementation is bottom-up and provides approximations of a programs success patterns . Goal-dependent information such as call patterns is obtained using a magic-set transformation. A novel compositional approach is applied so that call patterns for arbitrary goals are derived in a precise and efficient way.

Completely positive maps on the CCR-algebra

Given any operator on the testfunction space, the general form of the induced completely positive map of the C*-algebra of the canonical commutation relations is characterized.

Deriving polymorphic type dependencies for logic programs using multiple incarnations of prop

This paper illustrates the application of abstract compilation using multiple incarnations of the domain Prop in deriving type dependencies for logic programs. We illustrate how dependencies can be derived in the presence of both monomorphic and polymorphic type information. Type dependencies generalize the recently proposed notion of directional types as well as the more common notion of groundness dependencies. Directional types have proven useful in a number of applications such as in proving termination. These applications, however, are based on type declarations. The main contribution of this paper is in the simplicity in which non-trivial type dependencies are inferred using abstract compilation and by associating each type with an incarnation of Prop. We illustrate the use of a semantics for open logic programs in maintaining space efficient analyses. Time efficiency is also maintained due to approximation of the type domain in a boolean lattice calling on results of universal algebra.