Journal of Applied Logic最新文献

In the domain of decision theoretic planning, the factored framework (Factored Markov Decision Process, fmdp) has produced optimized algorithms using structured representations such as Decision Trees (Structured Value Iteration (svi), Structured Policy Iteration (spi)) or Algebraic Decision Diagrams (Stochastic Planning Using Decision Diagrams (spudd)). Since it may be difficult to elaborate the factored models used by these algorithms, the architecture sdyna, which combines learning and planning algorithms using structured representations, was introduced. However, the state-of-the-art algorithms for incremental learning, for structured decision theoretic planning or for reinforcement learning require the problem to be specified only with binary variables and/or use data structures that can be improved in term of compactness. In this paper, we propose to use Multi-Valued Decision Diagrams (mdds) as a more efficient data structure for the sdyna architecture and describe a planning algorithm and an incremental learning algorithm dedicated to this new structured representation. For both planning and learning algorithms, we experimentally show that they allow significant improvements in time, in compactness of the computed policy and of the learned model. We then analyzed the combination of these two algorithms in an efficient sdyna instance for simultaneous learning and planning using mdds.

Most Relevant Explanation (MRE) is an inference problem in Bayesian networks that finds the most relevant partial instantiation of target variables as an explanation for given evidence. It has been shown in recent literature that it addresses the overspecification problem of existing methods, such as MPE and MAP. In this paper, we propose a novel hierarchical beam search algorithm for solving MRE. The main idea is to use a second-level beam to limit the number of successors generated by the same parent so as to limit the similarity between the solutions in the first-level beam and result in a more diversified population. Three pruning criteria are also introduced to achieve further diversity. Empirical results show that the new algorithm outperforms local search and regular beam search.

Weighted model counting (WMC) is a well-known inference task on knowledge bases, and the basis for some of the most efficient techniques for probabilistic inference in graphical models. We introduce algebraic model counting (AMC), a generalization of WMC to a semiring structure that provides a unified view on a range of tasks and existing results. We show that AMC generalizes many well-known tasks in a variety of domains such as probabilistic inference, soft constraints and network and database analysis. Furthermore, we investigate AMC from a knowledge compilation perspective and show that all AMC tasks can be evaluated using sd-DNNF circuits, which are strictly more succinct, and thus more efficient to evaluate, than direct representations of sets of models. We identify further characteristics of AMC instances that allow for evaluation on even more succinct circuits.

Multiple iterated revision requires advanced belief revision techniques that are able to integrate several pieces of new information into epistemic states. A crucial feature of this kind of revision is that the multiple pieces of information should be dealt with separately. Previous works have proposed several independence postulates which should ensure this. In this paper, we argue, first, that these postulates are too strong as they may enforce beliefs without justification, and second, that they are not necessary to ensure the principal aim of multiple revision. Instead, principles of conditional preservation guarantee a suitable handling of sets of sentences under revision. We formalize such a principle for multiple propositional revision for ranking functions, and we propose some novel postulates for multiple iterated revision that are in line with AGM and the Darwiche & Pearl postulates. We show that just a few fundamental postulates are enough to cover major approaches to (multiple) iterated belief revision, and that independence in the sense of Thielscher, Jin, and Delgrande is optional. As a proof of concept, we present propositional c-revisions of ranking functions.

The continuous time Bayesian network (CTBN) is a probabilistic graphical model that enables reasoning about complex, interdependent, and continuous-time subsystems. The model uses nodes to denote subsystems and arcs to denote conditional dependence. This dependence manifests in how the dynamics of a subsystem changes based on the current states of its parents in the network. While the original CTBN definition allows users to specify the dynamics of how the system evolves, users might also want to place value expressions over the dynamics of the model in the form of performance functions. We formalize these performance functions for the CTBN and show how they can be factored in the same way as the network, allowing what we argue is a more intuitive and explicit representation. For cases in which a performance function must involve multiple nodes, we show how to augment the structure of the CTBN to account for the performance interaction while maintaining the factorization of a single performance function for each node. We introduce the notion of optimization for CTBNs, and show how a family of performance functions can be used as the evaluation criteria for a multi-objective optimization procedure.

An anomaly, or outlier, is an object exhibiting differences that suggest it belongs to an as-yet undefined class or category. Early detection of anomalies often proves of great importance because they may correspond to events such as fraud, spam, or device malfunctions. By automating the creation of a ranking or list of deviations, we can save time and decrease the cognitive overload of the individuals or groups responsible for responding to such events.

Over the years many anomaly and outlier metrics have been developed. In this paper we propose a clustering-based score ensembling method for outlier detection. Using benchmark datasets we evaluate quantitatively the robustness and accuracy of different ensemble strategies. We find that ensembling strategies offer only limited value for increasing overall performance, but provide robustness by negating the influence of severely underperforming models.

The modal characterization theorem by J. van Benthem characterizes classical modal logic as the bisimulation invariant fragment of first-order logic. In this paper, we prove a similar characterization theorem for intuitionistic modal logic. For this purpose we introduce the notion of modal asimulation as an analogue of bisimulations. The paper treats four different fragments of first-order logic induced by their respective versions of Kripke-style semantics for modal intuitionistic logic. It is shown further that this characterization can be easily carried over to arbitrary first-order definable subclasses of classical first-order models.

The paper exposes the relevance of permuting conversions (in natural-deduction systems) to the role of such systems in the theory of meaning known as proof-theoretic semantics, by relating permuting conversion to harmony, hitherto related to normalisation only. This is achieved by showing the connection of permuting conversion to the general notion of canonicity, once applied to arbitrary derivations from open assumption. In the course of exposing the relationship of permuting conversions to harmony, a general definition of the former is proposed, generalising the specific cases of disjunction and existential quantifiers considered in the literature.

A number of procedures for checking the satisfiability of formulas in the important branching time temporal logic CTL* have recently been proposed.

In this paper we develop techniques to support such reasoning based on rewrite rules. We consider how to automatically generate and quickly validate such rewrites. The techniques are based on use of so-called non-local and bundled variants of computation tree logics.

We show that such rules can quickly simplify CTL* formulas. These simplified formulas are shorter and easier to reason with using existing decision procedures for CTL*, as demonstrated by significant speed-ups across a wide range of benchmark formulas.

In this paper we provide both an Associative Commutative sound, confluent and terminating term rewrite system as well as a simple greedy rewrite system. We test them on a useful set of benchmarks. There are practical applications for theorem-proving and model-checking.

The article proposes a way to add marketing into the standard threshold model of social networks. Within this framework, the article studies logical properties of the influence relation between sets of agents in social networks. Two different forms of this relation are considered: one for promotional marketing and the other for preventive marketing. In each case a sound and complete logical system describing properties of the influence relation is proposed. Both systems could be viewed as extensions of Armstrong's axioms of functional dependency from the database theory.