We propose new mechanisms that can be used by a demand response aggregator to flexibly shift the charging of electric vehicles (EVs) to times where cheap but intermittent renewable energy is in high supply. Here, it is important to consider the constraints and preferences of EV owners, while eliminating the scope for strategic behaviour. To achieve this, we propose, for the first time, a generic class of incentive mechanisms for settings with both varying marginal electricity costs and multidimensional preferences. We show these are dominant strategy incentive compatible, i.e., EV owners are incentivised to report their constraints and preferences truthfully. We also detail a specific instance of this class, show that it achieves ≈98% of the optimal in realistic scenarios and demonstrate how it can be adapted to trade off efficiency with profit.

Existing techniques for inference in probabilistic logic programs are sequential: they first compute the relevant propositional formula for the query of interest, then compile it into a tractable target representation and finally, perform weighted model counting on the resulting representation. We propose Tp-compilation, a new inference technique based on forward reasoning. Tp-compilation proceeds incrementally in that it interleaves the knowledge compilation step for weighted model counting with forward reasoning on the logic program. This leads to a novel anytime algorithm that provides hard bounds on the inferred probabilities. Furthermore, an empirical evaluation shows that Tp-compilation effectively handles larger instances of complex real-world problems than current sequential approaches, both for exact and for anytime approximate inference.

Deep neural networks (DNNs) show very strong performance on many machine learning problems, but they are very sensitive to the setting of their hyperparameters. Automated hyperparameter optimization methods have recently been shown to yield settings competitive with those found by human experts, but their widespread adoption is hampered by the fact that they require more computational resources than human experts. Humans have one advantage: when they evaluate a poor hyperparameter setting they can quickly detect (after a few SGD steps) that the resulting network performs poorly and terminate the corresponding evaluation to save time. Here, we mimic this early termination of bad runs based on a probabilistic model that extrapolates performance from the first part of a learning curve. Experiments with different neural network architectures show that our resulting approach speeds up state-of-the-art hyperparameter optimization methods for DNNs roughly twofold, enabling them to find DNN settings that yield better performance than those chosen by human experts.

Most logic-based machine learning algorithms rely on an Occamist bias where textual complexity of hypotheses is minimised. Within Inductive Logic Programming (ILP), this approach fails to distinguish between the efficiencies of hypothesised programs, such as quick sort (O(n log n)) and bubble sort (O(n 2 )). This paper addresses this issue by considering techniques to minimise both the textual complexity and resource complexity of hypothesised robot strategies. We develop a general framework for the problem of minimising resource complexity and show that on two robot strategy problems, 1) Postman 2) Sorter (recursively sort letters for delivery), the theoretical resource complexities of optimal strategies vary depending on whether objects can be composed within a strategy. The approach considered is an extension of Meta-Interpretive Learning (MIL), a recently developed paradigm in ILP which supports predicate invention and the learning of recursive logic programs. We introduce a new MIL implementation, Metagol O , and prove its convergence, with increasing numbers of randomly chosen examples to optimal strategies of this kind. Our experiments show that Metagol O learns theoretically optimal robot sorting strategies, which is in agreement with the theoretical predictions showing a clear divergence in resource requirements as the number of objects grows. To the authors’ knowledge this paper is the first demonstration of a learning algorithm able to learn optimal resource complexity robot strategies and algorithms for sorting lists.

Context-aware systems use data collected at runtime to recognize certain predefined situations and trigger adaptations. This can be implemented using ontology-based data access (OBDA), which augments classical query answering in databases by adopting the open-world assumption and including domain knowledge provided by an ontology. We investigate temporalized OBDA w.r.t. ontologies formulated in EL, a description logic that allows for efficient reasoning and is successfully used in practice. We consider a recently proposed temporalized query language that combines conjunctive queries with the operators of propositional linear temporal logic (LTL), and study both data and combined complexity of query entailment in this setting. We also analyze the satisfiability problem in the similar formalism EL-LTL.

Product review sites such as TripAdvisor, Yelp or Amazon provide a single, non personalized ranking of products. The sparse review data makes personalizing recommendations difficult. Topic Profile Collaborative Filtering exploits review texts to identify user profiles as a basis for similarity. We show that careful use of the available data and separating users into classes can greatly improve the performance of such techniques. We significantly improve MAE, RMSE, and Kendall tau, compared to the previous best results. In addition, we show that personalization does not benefit all the users to the same extent. We propose switching between a personalized and a non personalized method based on the user opinion profile. We show that the user's opinionatedness is a good indicator of whether the personalization will work or not.

We introduce epistemic quantified boolean logic (EQBL), an extension of propositional epistemic logic with quantification over propositions. We show that EQBL can express relevant properties about agents’ knowledge in multi-agent contexts, such as “agent a knows as much as agent b”. We analyse the expressiveness of EQBL through a translation into monadic second-order logic, and provide completeness results w.r.t. various classes of Kripke frames. Finally, we prove that model checking EQBL is PSPACE-complete. Thus, the complexity of model checking EQBL is no harder than for (non-modal) quantified boolean logic.

Argumentation mining aims to automatically identify structured argument data from unstructured natural language text. This challenging, multi-faceted task is recently gaining a growing attention, especially due to its many potential applications. One particularly important aspect of argumentation mining is claim identification. Most of the current approaches are engineered to address specific domains. However, argumentative sentences are often characterized by common rhetorical structures, independently of the domain. We thus propose a method that exploits structured parsing information to detect claims without resorting to contextual information, and yet achieve a performance comparable to that of state-of-the-art methods that heavily rely on the context.

List of committee members on the program committees of the IJCAI 2015 conference.

We propose a versatile framework for combining knowledge bases in modular systems with preferences. In our formalism, each module (knowledge base) can be specified in a different language. We define the notion of a preference-based modular system that includes a formalization of meta-preferences. We prove that our formalism is robust in the sense that the operations for combining modules preserve the notion of  a preference-based modular system. Finally, we formally demonstrate correspondences between our framework and the related preference formalisms of cp-nets and preference-based planning. Our framework allows one to use  these preference formalisms (and others) in combination, in the same modular system.