In this paper, we propose a new framework for constructing text metrics which can be used to compare and support inferences among terms and sets of terms. Our metric is derived from data-driven kernels on graphs that let us capture global relations among terms and sets of terms, regardless of their complexity and size. To compute the metric efficiently for any two subsets of terms, we develop an approximation technique that relies on the precompiled term-term similarities. To scale-up the approach to problems with huge number of terms, we develop and experiment with a solution that subsamples the term space. We demonstrate the benefits of the whole framework on two text inference tasks: prediction of terms in the article from its abstract and query expansion in information retrieval.

Lifted message passing algorithms exploit repeated structure within a given graphical model to answer queries efficiently. Given evidence, they construct a lifted network of supernodes and superpotentials corresponding to sets of nodes and potentials that are indistinguishable given the evidence. Recently, efficient algorithms were presented for updating the structure of an existing lifted network with incremental changes to the evidence. In the inference stage, however, current algorithms need to construct a separate lifted network for each evidence case and run a modified message passing algorithm on each lifted network separately. Consequently, symmetries across the inference tasks are not exploited. In this paper, we present a novel lifted message passing technique that exploits symmetries across multiple evidence cases. The benefits of this multi-evidence lifted inference are shown for several important AI tasks such as computing personalized PageRanks and Kalman filters via multi-evidence lifted Gaussian belief propagation.

In this work we propose an approach for generalization in continuous domain Reinforcement Learning that, instead of using a single function approximator, tries many different function approximators in parallel, each one defined in a different region of the domain. Associated with each approximator is a relevance function that locally quantifies the quality of its approximation, so that, at each input point, the approximator with highest relevance can be selected. The relevance function is defined using parametric estimations of the variance of the q-values and the density of samples in the input space, which are used to quantify the accuracy and the confidence in the approximation, respectively. These parametric estimations are obtained from a probability density distribution represented as a Gaussian Mixture Model embedded in the input-output space of each approximator. In our experiments, the proposed approach required a lesser number of experiences for learning and produced more stable convergence profiles than when using a single function approximator.

Following the recent trend of studying the theory of belief revision under the Horn fragment of propo- sitional logic this paper develops a fully charac- terised Horn contraction which is analogous to the traditional transitively relational partial meet contraction [Alchourron et al., 1985]. This Horn con- traction extends the partial meet Horn contraction studied in [Delgrande and Wassermann, 2010] so that it is guided by a transitive relation that models the ordering of plausibility over sets of beliefs.

This paper focuses on computing first-order theories under either stable model semantics or circumscription. A reduction from first-order theories to logic programs under stable model semantics over finite structures is proposed, and an embedding of circumscription into stable model semantics is also given. Having such reduction and embedding, reasoning problems represented by first-order theories under these two semantics can then be handled by using existing answer set solvers. The effectiveness of this approach in computing hard problems beyond NP is demonstrated by some experiments.

The generation of route descriptions is a fundamental task of navigation systems. A particular problem in this context is to identify routes that can easily be described and processed by users. In this work, we present a framework for representing route networks with the qualitative information necessary to evaluate and optimize route descriptions with regard to ambiguities in them. We identify different agent models that differ in how agents are assumed to process route descriptions while navigating through route networks. Further, we analyze the computational complexity of matching route descriptions and paths in route networks in dependency of the agent model. Finally we empirically evaluate the influence of the agent model on the optimization and the processing of route instructions.

Carneades is a recently proposed formalism for structured argumentation with varying proof standards. An open question is its relation with Dung's seminal abstract approach to argumentation. In this paper the two formalisms are formally related by translating Carneades into ASPIC+, another recently proposed formalism for structured argumentation. Since ASPIC+ is defined to generate Dung-style abstract argumentation frameworks, this in effect translates Carneades graphs into abstract argumentation frameworks. It is proven that Carneades always induces a unique Dung extension, which is the same in all of Dung's semantics.

The Game Description Language is a high-level, rule-based formalisms for communicating the rules of arbitrary  games to general game-playing systems, whose challenging task is to learn to play previously unknown games without human intervention. Originally designed for deterministic games with complete information about the game state, the language was recently extended to include randomness and imperfect information. However, determining the extent to which this enhancement allows to describe truly arbitrary games was left as an open problem. We provide a positive answer to this question by relating the extended Game Description Language to the universal, mathematical concept of extensive-form games, proving that indeed just any such game can be described faithfully.

Consequence-based ontology reasoning procedures have so far been known only for Horn ontology languages. A difficulty in extending such procedures is that non-Horn axioms seem to require reasoning by case, which causes non-determinism in tableau-based procedures. In this paper we present a consequence-based procedure for ALCH that overcomes this difficulty by using rules similar to ordered resolution to deal with disjunctive axioms in a deterministic way; it retains all the favourable attributes of existing consequence-based procedures, such as goal-directed “one pass” classification, optimal worst-case complexity, and “pay-asyou- go” behaviour. Our preliminary empirical evaluation suggests that the procedure scales well to non-Horn ontologies.

We propose a new approach based on model relaxation to compute minimum-cardinality diagnoses of a (faulty) system: We obtain a relaxed model of the system by splitting nodes in the system and compile the abstraction of the relaxed model into DNNF. Abstraction is obtained by treating self-contained sub-systems called cones as single components. We then use a novel branch-and-bound search algorithm and compute the abstract minimum-cardinality diagnoses of the system, which are later refined hierarchically, in a careful manner, to get all minimum-cardinality diagnoses of the system. Experiments on ISCAS-85 benchmark circuits show that the new approach is faster than the previous state-of-the-art hierarchical approach, and scales to all circuits in the suite for the first time.