Showing papers in "Ai Communications in 2011"

Potassco: The Potsdam Answer Set Solving Collection

Abstract: This paper gives an overview of the open source project Potassco, the Potsdam Answer Set Solving Collection, bundling tools for Answer Set Programming developed at the University of Potsdam.

Open issues for normative multi-agent systems

Abstract: A challenging problem currently addressed in the multi-agent systems area is the development of open systems; which are characterized by the heterogeneity of their participants and the dynamic features of both their participants and their environment. The main feature of agents in these systems is autonomy. It is this autonomy that requires regulation, and norms are a solution for this. Norms represent a tool for achieving coordination and cooperation among the members of a society. They have been employed in the field of Artificial Intelligence as a formal specification of deontic statements aimed at regulating the actions of software agents and the interactions among them. This article gives an overview of the most relevant works on norms for multi-agent systems. This review considers open multi-agent systems challenges and points out the main open questions that remain in norm representation, reasoning, creation, and implementation.

Special issue on answer set programming

Abstract: Since its introduction, answer set programming (ASP) has been widely applied to various knowledgeintensive tasks and combinatorial search problems. ASP was found to be closely related to SAT, which has led to new methods of computing answer sets using SAT solvers and techniques adapted from SAT. While this has been the most studied relationship, the relationship of ASP to other computing paradigms, such as constraint satisfaction, quantified boolean formulas (QBF), or first-order logic (FOL) is also the subject of active research. The goal of the annual ASPOCP (acronym for ASP and Other Computing Paradigms) workshop is to facilitate the discussion about crossing the boundaries of current ASP techniques, in combination with or inspired by other computing paradigms. In 2010, the third edition of the ASPOCP workshop (ASPOCP’10) was held in Edinburgh, Scotland, as part of the International Conference on Logic Programming (ICLP) in the frame of the 5th Federated Logic Conference (FLoC). The program included one invited talk by Torsten Schaub and nine regular paper presentations. This special issue of AI Communications contains extended and carefully reviewed versions of select contributions to ASPOCP’10. The articles reflect the range of areas the ASPOCP workshop tries to touch; in particular articles contained in this volume address preferential reasoning, domain-specific heuristics, symmetry breaking, as well as grounding issues. The first article “Potassco: The Potsdam answer set solving collection” [5] is the contribution by our invited speaker and his team. It gives a thorough overview of the suite of ASP tools that are developed at the University of Potsdam. This family of ASP tools

Recommendation of reliable users, social networks and high-quality resources in a Social Internetworking System

Abstract: Social Internetworking Systems are a significantly emerging new reality; they group together some social networks and allow their users to share resources, to acquire opinions and, more in general, to interact, even if these users belong to different social networks and, therefore, did not previously know each other. In this context, owing to the huge dimension of existing social networks, the capability of a Social Internetworking System to provide its users with recommendations of reliable users and social networks, as well as of high-quality resources, is extremely relevant. In the past, user and resource recommendation has been investigated in the context of a single social network, whereas it has still received a little attention in the context of a Social Internetworking System, owing to the novelty of this phenomenon. For the same reason, social network recommendation has received an even less attention. In this paper we propose a trust-based approach to face these challenges. Specifically, we introduce a model to represent and handle trust and reputation in a Social Internetworking System and propose an approach that exploits these parameters to compute the reliability of a user or a social network, as well as the quality of a resource. These last measures are then exploited to perform recommendations.

Ontology-based semantic clustering

Abstract: This thesis presents novel measures to estimate the degree of semantic similarity between words using one or more knowledge sources. Several evaluations show that they improve the accuracy of related works. These measures have been applied to clustering to compute the similarity/distance between individuals described by textual attributes. Clustering results show that a proper interpretation of textual data at a semantic level improves the quality of the clusters and ease their interpretation.

Learning and using domain-specific heuristics in ASP solvers

Abstract: In spite of the improvements in the performance of many solvers for model-based languages, it is still possible for the search algorithm to focus on the wrong areas of the search space, preventing the solver from returning a solution in an acceptable amount of time. This prospect is a real concern e.g. in an industrial setting, where users typically expect consistent performance. To overcome this problem, we propose a framework that allows learning and using domain-specific heuristics in the solvers. The learning is done offline, on representative instances from the target domain, and the learned heuristics are then used for choice-point selection. In this paper we focus on Answer Set Programming (ASP) solvers. In our experiments, the introduction of domain-specific heuristics improved performance quite substantially on hard instances, and in particular made overall performance more consistent by reducing the number of cases in which the solver timed out.

Dynamic Magic Sets and super-coherent answer set programs

Abstract: For many practical applications of ASP, for instance data integration or planning, query answering is important, and therefore query optimization techniques for ASP are of great interest. Magic Sets are one of these techniques, originally defined for Datalog queries (ASP without disjunction and negation). Dynamic Magic Sets (DMS) are an extension of this technique, which has been proved to be sound and complete for query answering over ASP programs with stratified negation. A distinguishing feature of DMS is that the optimization can be exploited also during the non-deterministic phase of ASP engines. In particular, after some assumptions have been made during the computation, parts of the program may become irrelevant to a query under these assumptions. This allows for dynamic pruning of the search space, which may result in exponential performance gains. In this paper, the correctness of DMS is formally established and proved for brave and cautious reasoning over the class of super-coherent ASP programs (ASP sc programs). ASPsc programs guarantee consistency (i.e., have answer sets) when an arbitrary set of facts is added to them. This result generalizes the applicability of DMS, since the class of ASPsc programs is richer than ASP programs with stratified negation, and in particular includes all odd-cycle-free programs. DMS has been implemented as an extension of DLV, and the effectiveness of DMS for ASPsc programs is empirically confirmed by experimental results with this system.

Symmetry-breaking answer set solving

Abstract: We investigate the role of symmetry detection and symmetry breaking in answer set programming to eliminate symmetric parts of the search space and, thereby, simplify the solution process We reduce symmetry detection to a graph automorphism problem which allows us to extract symmetries of a logic program from the symmetries of the constructed coloured graph The correctness of our reduction is proven We also propose an encoding of symmetry-breaking constraints in terms of permutation cycles and use only generators in this process to implicitly represent symmetries with exponential compression These ideas are formulated as preprocessing and implemented in a completely automated flow that first detects symmetries from a given answer set program, adds symmetry-breaking constraints, and can be applied to any existing answer set solver We demonstrate computational impact on benchmarks versus direct application of the solver

Potassco: The Potsdam Answer Set Solving Collection

Abstract: This paper gives an overview of the open source project Potassco, the Potsdam Answer Set Solving Collection, bundling tools for Answer Set Programming developed at the University of Potsdam.

Toward a social web of intelligent things

Abstract: In this paper we introduce the notion of “Social Web of Intelligent Things” (SWIT hereafter) as a an evolution of both the “Web of Things” and “Smart Objects” paradigms. In a SWIT, things become entities capable of an intelligent and social behavior. On the one hand, things maintain and socialize knowledge and can interact and communicate with people; on the other, “social” networks of people and things arise as a result of this interaction. SWIT is an evolution of social media that goes beyond the desktop paradigm and is a way of bridging the gap between real-life and virtual experiences. Interactions between people and things happen naturally in real life, augmenting and enhancing people's experiences. We claim that SWITs are a challenging area of research and application for Artificial Intelligence in the context of the Future Internet. The paper provides a characterization of SWITs, discussing the ingredients that are needed to create a SWIT; it introduces a framework for building SWITs and briefly presents an instance we developed and tested with users. In the paper we claim that many areas of Artificial Intelligence can contribute to the is framework, which in turn can stimulate research in these areas and on their integration.

Finite domain and symbolic inference methods for extensions of first-order logic

Abstract: In this dissertation, we investigate various sorts of reasoning on finite structures and theories in the logic FO(·), a rich extension of classical logic with, amongst others, inductive definitions and aggregates. In particular, we study the tasks of constraint propagation, grounding, model revision, and debugging for FO(·).

Building cost-sensitive decision trees for medical applications

Abstract: This thesis presents strategies for cost-sensitive learning We have developed an algorithm for decision tree induction that considers various types of costs The main ones were attribute costs and misclassification costs Other costs included, for instance, the “risk”, that is a measure of how invasive the test is We applied our strategy to train and to evaluate cost-sensitive decision trees on medical data The resulting trees provided a better cost-effective solution for a given problem

A weighted CSP approach to cost-optimal planning

Abstract: For planning to come of age, plans must be judged by a measure of quality, such as the total cost of actions. This paper describes an optimal-cost planner which guarantees global optimality whenever the planning problem has a solution. We code the extraction of an optimal plan, from a planning graph with a fixed number k of levels, as a weighted constraint satisfaction problem (WCSP). The specific structure of the resulting WCSP means that a state-of-the-art exhaustive solver was able to find an optimal plan in planning graphs containing several thousand nodes. Thorough experimental investigations demonstrated that using the planning graph in optimal planning is a practical possibility for problems of moderate size, although not competitive, in terms of computation time, with optimal state-space-search planners. Our general conclusion is, therefore, that planning-graph-based optimal planning is not the most efficient method for cost-optimal planning. Nonetheless, the notions of indispensable (sets of) actions and too-costly actions introduced in this paper have various potential applications in optimal planning. These actions can be detected very rapidly by analysis of the relaxed planning graph.

A logical characterisation of ordered disjunction

Abstract: In this paper we consider a logical treatment for the ordered disjunction operator × introduced by Brewka, Niemela and Syrjanen in their Logic Programs with Ordered Disjunctions (LPOD). LPODs are used to represent preferences in logic programming under the answer set semantics. Their semantics is defined by first translating the LPOD into a set of normal programs (called split programs) and then imposing a preference relation among the answer sets of these split programs. We concentrate on the first step and show how a suitable translation of the ordered disjunction as a derived operator into the logic of Here-and-There allows capturing the answer sets of the split programs in a direct way. We use this characterisation not only for providing an alternative implementation for LPODs, but also for checking several properties (under strongly equivalent transformations) of the × operator, like for instance, its distributivity with respect to conjunction or regular disjunction. We also make a comparison to an extension proposed by Karger, Lopes, Olmedilla and Polleres, that combines × with regular disjunction.

Finitely recursive programs: Decidability and bottom-up computation

Abstract: The support for function symbols in logic programming under answer set semantics allows us to overcome some modeling limitations of traditional Answer Set Programming (ASP) systems, such as the inability of handling infinite domains. On the other hand, admitting function symbols in ASP makes inference undecidable in the general case. Recently, the research community has been focusing on finding proper subclasses of programs with functions for which decidability of inference is guaranteed. The two major proposals, so far, are finitary programs and finitely-ground programs. These two proposals are somehow complementary: indeed, the former is conceived to allow decidable querying (by means of a top-down evaluation strategy), while the latter supports the computability of answer sets (by means of a bottom-up evaluation strategy). One of the main advantages of finitely-ground programs is that they can be “directly” evaluated by current ASP systems, which are based on a bottom-up computational model. However, there are also some interesting programs which are suitable for top-down query evaluation; but they do not fall in the class of finitely-ground programs. In this paper, we focus on disjunctive finitely recursive positive (DFRP) programs. We design a version of the magic sets technique for DFRP programs, which ensures query equivalence under both brave and cautious reasoning. We show that, if the input program is DFRP, then its magic-sets rewriting is guaranteed to be finitely ground. Reasoning on DFRP programs turns out to be decidable; and we provide also an effective method that allows one to simply perform this reasoning by using the ASP system DLV.

Max-SAT formalisms with hard and soft constraints

Abstract: The study of Max-SAT formalisms and the development of fast Max-SAT solvers have become very active research topics in the last few years. This PhD dissertation focus on the design, implementation and evaluation of solving techniques for two Max-SAT formalisms that incorporate the notion of partiality. The first one, called Block-SAT, is a new formalism that deals with blocks of clauses instead of individual clauses. The second one is Partial Max-SAT, which has become a standard in the community. For each formalism, we compare our solvers with state-of-the-art solvers using an exhaustive test suite. The empirical results provide evidence that the proposed solving techniques are competitive and often produce important speed-ups.

LTL model-checking for security protocols

Abstract: This thesis is about the application of automated reasoning techniques to the formal analysis of security protocols. More in detail, it proposes a general model-checking framework for security protocols based on a set-rewriting formalism that, coupled with the use of Linear Temporal Logic, allows for the specification of assumptions on principals and communication channels as well as complex security properties that are normally not handled by state-of-the-art protocol analyzers. The approach successfully combines encoding techniques originally developed for planning with bounded model-checking techniques. The effectiveness of the approach proposed is assessed against the formal analysis of relevant security protocols, with the detection of a severe security flaw in Google's SAML-based SSO for Google Apps and a previously unknown attack on a patched version of the ASW contract-signing protocol.

Constraint-based methods for scheduling discretionary services

Abstract: A project network composed of discretionary tasks typically exists in service professions, such as journalism, clinic, software development or financial analysis, where the quality (or value) of a task increases with the time spent on it. Since a longer task duration consumes more resources (i.e., workers' time), the project manager must strike a balance between quality and time by scheduling tasks and setting their durations while respecting the project deadline, precedence and resource constraints. We formulate this problem, give a polynomial-time optimal algorithm for the single capacity case and prove the NP-completeness of the general multiple capacity case. Then we develop two hybrid solution procedures integrating linear optimization and an AI search procedure - precedence constraint posting - for the general case. Our results verify the effectiveness of these procedures and show there exists a potential synergy between objectives of maintaining temporal flexibility and maximizing quality, which implies that existing techniques in building flexible schedules can be adapted to solve this new class of problems.

Efficient model building in competent genetic algorithms using DSM clustering

Abstract: Detecting multivariate interactions between the variables of a problem is a challenge in traditional genetic algorithms (GAs). This issue has been addressed in the literature as the linkage learning problem. It is widely acknowledged that the success of GA in solving any problem depends on the proper detection of multivariate interactions in the problem. Different approaches have thus been proposed to detect and represent such interactions. Estimation of distribution algorithms (EDAs) are amongst these approaches that have been successfully applied to a wide range of hard optimization problems. They build a model of the problem to detect multivariate interactions, but the model building process is often computationally intensive. In this paper, we propose a new clustering algorithm that turns pair-wise interactions in a dependency structure matrix (DSM) into an interaction model efficiently. The model building process is carried out before the evolutionary algorithm to save computational burden. The accurate interaction model obtained in this way is then used to perform an effective recombination of building blocks (BBs) in the GA. We applied the proposed approach to solve exemplar hard optimization problems with different types of linkages to show the effectiveness and efficiency of the proposed approach. Theoretical analysis and experiments showed that the building of an accurate model requires O(nlog (n)) number of fitness evaluations. The comparison of the proposed approach with some existing algorithms revealed that the efficiency of the model building process is enhanced significantly.

Partially decentralized reinforcement learning in finite, multi-agent Markov decision processes

Abstract: In this paper, we propose a novel, partially decentralized learning algorithm for the control of finite, multi-agent Markov Decision Process with unknown transition probabilities and reward values. One learning automaton is associated with each agent acting in a state and the automata acting within a state may communicate with each other. However, there is no communication between the automata present in different states, thus making the system partially decentralized. We propose novel algorithms so that the entire automata team converges to the policy that maximizes the long-term expected reward per step. Simulation results are presented to demonstrate the usefulness of the proposed algorithms.

Clustering preprocessing to improve time series forecasting

Abstract: This work proposes a novel general-purpose forecasting algorithm. It first extracts patterns from time series using the information provided by certain clustering techniques, which are applied as a first step of the approach. Moreover, the occurrence of data with especially unexpected values (outliers) is also addressed in this work. To deal with these outliers, a new hybrid methodology has been proposed, by inserting and adapting an existing approach based on the discovery of frequent episodes in sequences in the general scheme of prediction.

Graph-based data mining for biological applications

Abstract: In many real-world problems, one deals with input or output data that are structured. This thesis investigates the use of graphs as a representation for structured data and introduces relational learning techniques that can efficiently process them. We apply the techniques to two biological problems. On the one hand, we use decision trees to predict the functions of genes, of which the hierarchical relationships can be structured as a graph. On the other hand, we predict chemical activity of molecules by representing them as graphs. We show that, by exploiting graph properties, efficient learning techniques can be developed. It turns out that in both cases, the relational models are not only learned more efficiently, but their predictive performance significantly improves as well.

Foundations and applications of generalized planning

Abstract: Generalized planning problems capture planning problems with uncertainties in object quantities and properties. While the generalized planning problem is incomputable in general, this thesis presents a new formalization of the problem, a study of the limits of computability of generalized plans and algorithms for solving a broad class of generalized planning problems.

Exploiting multidimensional data for web site automation

Abstract: The continuous growth in size and usage of the World Wide Web poses a number of challenging research problems. This thesis investigates the use of multidimensional data to automate the personalization and management activities of web sites. We pay particular attention to the use of additional dimensions (e.g., contextual or background information) in traditional user-item based top-N recommender systems and propose a multidimensional approach for this purpose. To support our research, we also propose a data warehouse to collect and compile information regarding the activity on a web site in terms of usage, content and structure. We show that, by exploiting multidimensional data, efficient web automation methods can be developed and used to improve the personalization and management of web sites.

Symmetry-breaking answer set solving

Abstract: We investigate the role of symmetry detection and symmetry breaking in answer set programming to eliminate symmetric parts of the search space and, thereby, simplify the solution process. We reduce...

The bare necessities: Adaptive assistance for wheelchair control

Abstract: In assisted wheelchair navigation it is important to optimize help: an excess may lead to loss of residual skills, whereas a lack may lead to failure in completing a trajectory. Help needed depends on diagnosis, condition, task and environment. This dissertation presents a new approach to personalize assistance through learning. It has been successfully tested and validated at a rehabilitation hospital.

Implicit abstraction heuristics for cost-optimal planning

Abstract: State-space search with explicit abstraction heuristics is a state of the art approach to cost-optimal planning. These heuristics, however, have the limitation that the size of the abstract space must be bounded by some constant. We therefore introduce the notion of (additive) implicit abstractions, in which the planning task is abstracted by instances of tractable fragments of cost-optimal planning. We show that the fork-decomposition, a concrete instance of this framework based on two novel such fragments, compares favorably to the state of the art in cost-optimal planning. Additive ensembles of admissible heuristics are used in cost-optimal planning to exploit the individual strengths of different admissible heuristics. Continuing our focus on abstraction heuristics, we describe a procedure that takes a planning problem, a search state, and a set of admissible heuristics, and derives an optimal additive composition of these heuristics with respect to the given state. We show that this procedure is polynomial-time for arbitrary sets of abstraction heuristics.

On the use of machine learning methods for modern drug discovery

Abstract: Machine learning techniques provide a strong basis for enhancing the process of drug discovery within the pharmaceutical industry. During the last decade, chemoinformatics - briefly defined as informatics applied on chemical data - has gained much importance due to the economical benefits obtained from the application of in silico (i.e., computer-based) models. Prediction of candidate compounds for medicinal use is a hard task due to the complex and usually unknown relationships between structure and biological properties. This short article aims at summarizing the main contributions of a PhD thesis [PhD thesis, Universidad Nacional del Sur, Argentina, 2010] and, at the same time, at encouraging research on this challenging area.

Planning with rich goals, preferences and procedural operators via reformulation

Abstract: We summarize the main contributions of the thesis “Effective search techniques for non-classical planning via reformulation”, which received an Honorable Mention for the ICAPS Best Dissertation Award, 2011. The thesis explored how to leverage state-of-the-art classical planning techniques in order to solve compelling non-classical planning problems such as planning with temporally extended goals and preferences, planning with procedural control and planning with procedural operators.