International Journal on Artificial Intelligence Tools

This study introduces an advanced Combinatorial Reverse Auction (CRA), multi-units, multiattributes and multi-objective, which is subject to buyer and seller trading constraints. Conflicting objectives may occur since the buyer can maximize some attributes and minimize some others. To address the Winner Determination (WD) problem for this type of CRAs, we propose an optimization approach based on genetic algorithms that we integrate with our variants of diversity and elitism strategies to improve the solution quality. Moreover, by maximizing the buyer’s revenue, our approach is able to return the best solution for our complex WD problem. We conduct a case study as well as simulated testing to illustrate the importance of the diversity and elitism schemes. We also validate the proposed WD method through simulated experiments by generating large instances of our CRA problem. The experimental results demonstrate on one hand the performance of our WD method in terms of several quality measures, like solution quality, run-time complexity and trade-off between convergence and diversity, and on the other hand, it’s significant superiority to well-known heuristic and exact WD techniques that have been implemented for much simpler CRAs.

User-generated content such as comments or tweets (also called by social information) following a Web document provides additional information for enriching the content of an event mentioned in sentences. This paper presents a framework named SoSVMRank, which integrates the user-generated content of a Web document to generate a highquality summarization. In order to do that, the summarization was formulated as a learning to rank task, in which comments or tweets are exploited to support sentences in a mutual reinforcement fashion. To model sentence-comment (or tweet) relation, a set of local and social features are proposed. After ranking, top m ranked sentences and comments (or tweets) are selected as the summarization. To validate the efficiency of our framework, sentence and story highlight extraction tasks were taken as a case study on three datasets in two languages, English and Vietnamese. Experimental results indicate that: (i) our new features improve the summary performance of the framework in term of ROUGE-scores compared to state-of-the-art baselines and (ii) the integration of user-generated content benefits single-document summarization.

Multiagent patrolling is the problem faced by a set of agents that have to visit a set of sites to prevent or detect some threats or illegal actions. Although it is commonly assumed that patrollers share a common objective, the issue of cooperation between the patrollers has received little attention. Over the last years, the focus has been put on patrolling strategies to prevent a one-shot attack from an adversary. This adversary is usually assumed to be fully rational and to have full observability of the system. Most approaches are then based on game theory and consists in computing a best response strategy. Nonetheless, when patrolling frontiers, detecting illegal fishing or poaching; patrollers face multiple adversaries with limited observability and rationality. Moreover, adversaries can perform multiple illegal actions over time and space and may change their strategies as time passes. In this paper, we propose a multiagent planning approach that enables effective cooperation between a team of patrollers in uncertain environments. Patrolling agents are assumed to have partial observability of the system. Our approach allows the patrollers to learn a generic and stochastic model of the adversaries based on the history of observations. A wide variety of adversaries can thus be considered with strategies ranging from random behaviors to fully rational and informed behaviors. We show that the multiagent planning problem can be formalized by a non-stationary DEC- POMDP. In order to deal with the non-stationary, we introduce the notion of context. We then describe an evolutionary algorithm to compute patrolling strategies on-line, and we propose methods to improve the patrollers’ performance.

This research uses Machine Learning (ML) techniques in order to aid search-based (SB) algorithms to improve level generation for physics-based puzzle games. These algorithms’ performance are improved by reducing simulation time when the ML techniques are applied. Classification algorithms prevent levels with undesired traits to be evaluated during the simulation phase of the SB procedures. An Angry Birds clone is used for conducting the experiments and results report improvement using the combined approach against every SB algorithm by itself.

Manual creation of machine learning ensembles is a hard and tedious task which requires an expert and a lot of time. In this work we describe a new version of the GP-ML algorithm which uses genetic programming to create machine learning workows (combinations of preprocessing, classification, and ensembles) automatically, using strongly typed genetic programming and asynchronous evolution. The current version improves the way in which the individuals in the genetic programming are created and allows for much larger workows. Additionally, we added new machine learning methods. The algorithm is compared to the grid search of the base methods and to its previous versions on a set of problems from the UCI machine learning repository.

Many Artificial Intelligence techniques have been developed for intelligent and autonomous systems to act and make rational decisions based on perceptions of the world state. Among these techniques, HTN (Hierarchical Task Network) planning is one of the most used in practice. HTN planning is based on expressive languages allowing to specify complex expert knowledge for real world domains. At the same time, many preprocessing techniques for classical planning were proposed to speed up the search. One of these technique, named grounding, consists in enumerating and instantiating all the possible actions from the planning problem descriptions. This technique has proven its effectiveness. Therefore, combining the expressiveness of HTN planning with the efficiency of the grounding preprocessing techniques used in classical planning is a very challenging issue. In this paper, we propose a generic algorithm to ground the domain representation for HTN planning. We show experimentally that grounding process improves the performances of state of the art HTN planners on a range of planning problems from the International Planning Competition (IPC).

Nowadays, both public and private organizations own large private text-based data repositories with critical information. The information stored in these data silos is usually queried through information retrieval systems based on indexes, which yield hundreds or thousands of results when interrogated using keywords. In order to improve data accessibility when searching for specific information, the use of infoboxes can be very useful. The generation such infoboxes is by itself a complex problem, but in this type of isolated environments, it becomes even harder as the selection of the entities and their attributes can be conditioned by local and very specific parameters. In this work, we propose a methodology to tackle this special problem, combining classical approaches with machine learning, and leveraging the resources provided by the Semantic Web. The working methodology has been applied to two well-known datasets, and also it has been tested on a real environment scenario, showing the feasibility of our approach.

Personality prediction based on textual data is one topic gaining attention recently for its potential in large-scale personalized applications such as social media-based marketing. However, when applying this technology in real-world applications, users often encounter situations in which the personality traits derived from different sources (e.g., social media posts versus emails) are inconsistent. Varying results for the same individual renders the technology ineffective and untrustworthy. In this paper, we demonstrate the impact of domain differences in automated text-based personality prediction. We also propose different approaches for domain error correction to meet different needs: (a) single or multi-domain correction and (b) outcome-based or input feature-based error correction. We conduct comprehensive experiments to evaluate the effectiveness of these methods. Our findings demonstrate a significant improvement of prediction accuracy with the proposed methods. (e.g., 20–30% relative error reduction using outcome-based error correction or 48% increase of F1 score using feature-based error correction).

Computing a (Union of Conjunctive Queries — UCQ) rewriting ℛ for an input query and ontology and evaluating it over the given dataset is a prominent approach to query answering over ontologies. However, ℛ can be large and complex in structure hence additional techniques, like query subsumption and data constraints, need to be employed in order to minimize ℛ and lead to an efficient evaluation. Although sound in theory, how to efficiently and effectively implement many of these techniques in practice could be challenging. For example, many systems do not implement query subsumption. In the current paper we present several practical techniques for UCQ rewriting minimization. First, we present an optimized algorithm for eliminating redundant (w.r.t. subsumption) queries as well as a novel framework for rewriting minimization using data constraints. Second, we show how these techniques can also be used to speed up the computation of ℛ in first place. Third, we integrated all our techniques in our query rewriting system IQAROS and conducted an extensive experimental evaluation using many artificial as well as challenging real-world ontologies obtaining encouraging results as, in the vast majority of cases, our system is more efficient compared to the two most popular state-of-the-art systems.