Browsing
     by title


0-9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

or enter first few letters:   
OK
Full Text
Peer Reviewed
See detailLearning movement patterns in mobile networks: a generic method
François, Jean-Marc; Leduc, Guy ULg; Martin, Sylvain ULg

in European Wireless 2004 (2004, February)

Predicting terminals movements in mobile networks is useful for more than one reason, in particular for routing management. A way to do such prediction is to learn the movement patterns of mobile nodes ... [more ▼]

Predicting terminals movements in mobile networks is useful for more than one reason, in particular for routing management. A way to do such prediction is to learn the movement patterns of mobile nodes passing by an access router. In this paper, the information (e.g. layer 2 measurements) related to the different paths followed by mobiles are learned using a hidden Markov model. Simulations have been done using this method and show it can handle different layer~2 signals and collect statistical information when no such signal is available. Furthermore, the method works when no information is available and can be extended so as to guess the timing of the handoffs. [less ▲]

Detailed reference viewed: 68 (6 ULg)
Peer Reviewed
See detailLearning new names for new stimuli : Making the connection
Thi, Jean-Pierre; Comblain, Annick ULg

Poster (1999, September)

Detailed reference viewed: 3 (1 ULg)
Peer Reviewed
See detailLearning Object-specific Grasp Affordance Densities
Detry, Renaud ULg; Başeski, Emre; Krüger, Norbert et al

in International Conference on Development and Learning (2009)

Detailed reference viewed: 13 (1 ULg)
Peer Reviewed
See detailLearning Objects and Grasp Affordances through Autonomous Exploration
Kraft, Dirk; Detry, Renaud ULg; Pugeault, Nicolas et al

in International Conference on Computer Vision Systems (2009)

Detailed reference viewed: 10 (3 ULg)
See detailLearning of Multi-Dimensional, Multi-Modal Features for Robotic Grasping
Detry, Renaud ULg

Doctoral thesis (2010)

Detailed reference viewed: 14 (1 ULg)
Full Text
Peer Reviewed
See detailLearning parameters in discrete naive Bayes models by computing fibers of the parametrization map
Auvray, Vincent; Wehenkel, Louis ULg

in NIPS ´08 Workshop: Algebraic and combinatorial methods in machine learning (2008, December 20)

Discrete Naive Bayes models are usually defined parametrically with a map from a parameter space to a probability distribution space. First, we present two families of algorithms that compute the set of ... [more ▼]

Discrete Naive Bayes models are usually defined parametrically with a map from a parameter space to a probability distribution space. First, we present two families of algorithms that compute the set of parameters mapped to a given discrete Naive Bayes distribution satisfying certain technical assumptions. Using these results, we then present two families of parameter learning algorithms that operate by projecting the distribution of observed relative frequencies in a dataset onto the discrete Naive Bayes model considered. They have nice convergence properties, but their computational complexity grows very quickly with the number of hidden classes of the model. [less ▲]

Detailed reference viewed: 12 (1 ULg)
Full Text
Peer Reviewed
See detailLearning Probabilistic Discriminative Models of Grasp Affordances under Limited Supervision
Erkan, Ayse; Kroemer, Oliver; Detry, Renaud ULg et al

in IEEE/RSJ International Conference on Intelligent Robots and Systems (2010)

Detailed reference viewed: 19 (5 ULg)
Full Text
Peer Reviewed
See detailLearning Reliability Models of Grid Resource Supplying
Briquet, Cyril ULg; de Marneffe, Pierre-Arnoul ULg

in Bubak, Marian; Turala, Michal; Wiatr, Kazimierz (Eds.) CGW'05 Proceedings (2005, November 22)

Resource exchange between Grid participants is at the core of Grid computing. Distributed bartering is a distributed and moneyless method of resource exchange. Recent work related to distributed bartering ... [more ▼]

Resource exchange between Grid participants is at the core of Grid computing. Distributed bartering is a distributed and moneyless method of resource exchange. Recent work related to distributed bartering has mainly dealt with resource supplying. However, Grid participants still face an unstable resource environment due to the partial and intermittent nature of the exchanged resources. The problem considered in this paper is the unreliability of resource supplying. Though it cannot be totally avoided, a proactive stance may lower its impact in the long run. We propose to explore the reduction of performance variability by improving resource consumption. The goal is to enable Grid participants to identify and avoid unreliable resource suppliers by learning reliability models of resource supplying. A Machine Learning problem is defined and the generated models are applied to select more reliable resources in the hope of improving resource consumption. [less ▲]

Detailed reference viewed: 26 (6 ULg)
Peer Reviewed
See detailLearning subordinate and basic level category names : the misssing link
Thibaut, Jean-Pierre; Comblain, Annick ULg; Laurent, Marie et al

Poster (1999, December)

Chez les enfants normaux, l'acquisition du lexique est soumise à quatre principes de base acquis dans un ordre développemental déterminé : le principe de l'objet total, le principe d'extensibilité, le ... [more ▼]

Chez les enfants normaux, l'acquisition du lexique est soumise à quatre principes de base acquis dans un ordre développemental déterminé : le principe de l'objet total, le principe d'extensibilité, le principe taxinomique et le principe d'appariement d'un nouveau nom à une catégorie sans nom. Les enfants présentant un retard mental, du moins les enfants atteints des syndromes de Down et de Williams-Beuren) sont également soumis à l'application de ces principes lors de l'acquisition du lexique. Si le cadre de travail fourni par ces quatre principes d'acquisition du lexique permet de dégager des pistes d'intervention utilisables avec les jeunes enfants, des recherches sont encore nécessaires afin d'établir le type d'activités le plus efficace dans le cadre d'une stimulation lexicale chez l'enfant normal et à fortiori chez l'enfant atteint d'un retard mental. Il est primordial de garder à l'esprit que le but d'une intervention est d'optimaliser les bases cognitives et linguistiques sous-tendant un principe particulier et non d'enseigner à l'enfant une stratégie de surface ne faisant que mimer le principe (Mervis & Bertrand, 1993). [less ▲]

Detailed reference viewed: 26 (0 ULg)
Peer Reviewed
See detailLearning Tactile Characterizations Of Object- And Pose-specific Grasps
Bekiroglu, Yasemin; Detry, Renaud ULg; Kragic, Danica

in IEEE/RSJ International Conference on Intelligent Robots and Systems (2011)

Detailed reference viewed: 9 (0 ULg)
Full Text
Peer Reviewed
See detailLearning to Learn: Assessment of Metacognitive competencies
Frenkel, Stéphanie ULg; Nobile, Debora ULg

Poster (2014, March)

Facing the rise in requests for consultations related to school learning difficulties as well as requests for tools and additional methods by professional field workers, the EDUCA + project constitutes ... [more ▼]

Facing the rise in requests for consultations related to school learning difficulties as well as requests for tools and additional methods by professional field workers, the EDUCA + project constitutes one of the possible answers. This presentation focuses on one of the project’s goals, which is the creation of a Dynamic Assessment (DA) tailored to primary and secondary school learners. In this perspective, a brief introduction to the EDUCA + project will first be given, then, DA itself will be addressed along three lines: a reminder of its main characteristics, a review of the literature on existing tests, and the variables to evaluate. Finally, the planned perspectives will be presented. [less ▲]

Detailed reference viewed: 22 (5 ULg)
Peer Reviewed
See detailLearning to Learn: Assessment of Metacognitive competencies
Frenkel, Stéphanie ULg; Nobile, Debora ULg

in IATED (Ed.) INTED2014 Proceedings (2014, March)

Numerous students are having school difficulties linked to the way they learn. Some authors speak of a “metacognitive deficit”. We refer to a “sleeping potential” instead (Frenkel, 2013, in press; Frenkel ... [more ▼]

Numerous students are having school difficulties linked to the way they learn. Some authors speak of a “metacognitive deficit”. We refer to a “sleeping potential” instead (Frenkel, 2013, in press; Frenkel & Deforge, in press; Frenkel & Nobile, 2013). Be it psychologists, teachers or parents, all wish to develop their skills in order to help these students. This is the case in primary and secondary school. Metacognitive abilities play a central role in learning (e.g., Frenkel & Deforge, in press; Giasson, 2001; Grangeat, 1997; Hessels & Hessels-Schlatter, 2010b; Lumbelli, 2003; Poissant, Poëllhuber & Falardeau, 1994; Rozencwajg, 2003; Veenman, Kok & Blöte, 2005) and thus in successful school learning (Büchel, 2013a, 2013b; Van der Stel & Veenman, 2010; Wang, Haertel & Walberg, 1994). However, studying them requires that we clarify what is meant by “metacognition” and “metacognitive abilities”. In this framework, we developed the EDUCA + project which is intended to provide possible solutions. EDUCA + is based on wide field experience. Its theoretical background is based on a substantial review of the literature. Its objective is to increase the expertise of “front line” field workers by developing specific products such as tools, training courses, services, and a website (Frenkel, in press). This will notably enable them to develop their expertise, detect « sleeping » potential, diagnose, give advice when necessary, intervene (prevention and remediation) and use the tools efficiently. Two types of tools are being designed. On the one hand, assessment tools (tests allowing to put forward the learner’s strengths and weaknesses as well as the scope of his/her “sleeping potential”). On the other hand, intervention tools (short prevention vs. remediation programs). This also includes training courses and services. The creation of a website also aims to reinforce the actions of EDUCA + (personalized access depending on the internaut’s profile: students, parents, professionals). The aim of this paper is to present this research project and its main theoretical background. [less ▲]

Detailed reference viewed: 25 (11 ULg)
Full Text
Peer Reviewed
See detailLearning to play K-armed bandit problems
Maes, Francis ULg; Wehenkel, Louis ULg; Ernst, Damien ULg

in Proceedings of the 4th International Conference on Agents and Artificial Intelligence (ICAART 2012) (2012, February)

We propose a learning approach to pre-compute K-armed bandit playing policies by exploiting prior information describing the class of problems targeted by the player. Our algorithm first samples a set of K ... [more ▼]

We propose a learning approach to pre-compute K-armed bandit playing policies by exploiting prior information describing the class of problems targeted by the player. Our algorithm first samples a set of K-armed bandit problems from the given prior, and then chooses in a space of candidate policies one that gives the best average performances over these problems. The candidate policies use an index for ranking the arms and pick at each play the arm with the highest index; the index for each arm is computed in the form of a linear combination of features describing the history of plays (e.g., number of draws, average reward, variance of rewards and higher order moments), and an estimation of distribution algorithm is used to determine its optimal parameters in the form of feature weights. We carry out simulations in the case where the prior assumes a fixed number of Bernoulli arms, a fixed horizon, and uniformly distributed parameters of the Bernoulli arms. These simulations show that learned strategies perform very well with respect to several other strategies previously proposed in the literature (UCB1, UCB2, UCB-V, KL-UCB and $\epsilon_n$-GREEDY); they also highlight the robustness of these strategies with respect to wrong prior information. [less ▲]

Detailed reference viewed: 94 (17 ULg)
Full Text
See detailLearning to Predict End-to-End Network Performance
Liao, Yongjun ULg

Doctoral thesis (2013)

The knowledge of end-to-end network performance is essential to many Internet applications and systems including traffic engineering, content distribution networks, overlay routing, application-level ... [more ▼]

The knowledge of end-to-end network performance is essential to many Internet applications and systems including traffic engineering, content distribution networks, overlay routing, application-level multicast, and peer-to-peer applications. On the one hand, such knowledge allows service providers to adjust their services according to the dynamic network conditions. On the other hand, as many systems are flexible in choosing their communication paths and targets, knowing network performance enables to optimize services by e.g. intelligent path selection. In the networking field, end-to-end network performance refers to some property of a network path measured by various metrics such as round-trip time (RTT), available bandwidth (ABW) and packet loss rate (PLR). While much progress has been made in network measurement, a main challenge in the acquisition of network performance on large-scale networks is the quadratical growth of the measurement overheads with respect to the number of network nodes, which renders the active probing of all paths infeasible. Thus, a natural idea is to measure a small set of paths and then predict the others where there are no direct measurements. This understanding has motivated numerous research on approaches to network performance prediction. Commonly, the success of a prediction system is built on its scalability, efficiency, accuracy and practicability. For network performance prediction, two specific requirements have to be met. First, the prediction system should have a decentralized architecture which allows the natural deployment of the system within a networked application. Second, as different performance metrics are useful for different applications, the prediction system should be general and flexible to deal with various metrics in a unified framework. This thesis presents practical approaches to network performance prediction. There are three main contributions. First, the problem of network performance prediction is formulated as a matrix completion problem where the matrix contains performance measures between network nodes with some of them known and the others unknown and thus to be filled. This new formulation is advantageous in that it is flexible to deal with various metrics in a unified framework, despite their diverse nature. The only requirement is that the matrix to be completed has a low-rank characteristic, which has long been observed in performance matrices constructed from various networks and in various metrics. Second, the matrix completion problem is solved by a novel approach called Decentralized Matrix Factorization by Stochastic Gradient Descent (DMFSGD). The approach requires neither explicit constructions of matrices nor special nodes such as landmarks and central servers. Instead, by letting network nodes exchange messages with each other, matrix factorization is collaboratively and iteratively achieved at all nodes, with each node equally retrieving a number of measurements. The approach is practical in that it is simple, with no infrastructure, and is computationally lightweight, containing only vector operations. Third, instead of the conventional representation of exact metric values, this thesis also investigates coarse performance representations including binary classes (The performance is classified into binary classes of either ``good'' or ``bad''.) and ordinal ratings (The performance is quantized from 1 star to 5 stars.). Such more qualitative than quantitative measures not only fulfill the requirements of many Internet applications, but also reduce the measurement cost and enable a unified treatment of various metrics. In addition, as both class and rating measures can be nicely integrated in the matrix completion framework, the same DMFSGD approach is applicable for their prediction, with little modification required. The resulting prediction system has been extensively evaluated on various publicly-available datasets of two kinds of metrics, namely RTT and ABW. These experiments demonstrate not only the scalability and the accuracy of the DMFSGD approach but also its usability in real Internet applications. In addition, the benefits of predicting performance classes and ratings, rather than their actual values, are demonstrated by a case study on peer selection, a function that is commonly required in a number of network applications. [less ▲]

Detailed reference viewed: 33 (12 ULg)
Full Text
Peer Reviewed
See detailLearning to rank with extremely randomized trees
Geurts, Pierre ULg; Louppe, Gilles ULg

in JMLR: Workshop and Conference Proceedings (2011, January), 14

In this paper, we report on our experiments on the Yahoo! Labs Learning to Rank challenge organized in the context of the 23rd International Conference of Machine Learning (ICML 2010). We competed in both ... [more ▼]

In this paper, we report on our experiments on the Yahoo! Labs Learning to Rank challenge organized in the context of the 23rd International Conference of Machine Learning (ICML 2010). We competed in both the learning to rank and the transfer learning tracks of the challenge with several tree-based ensemble methods, including Tree Bagging, Random Forests, and Extremely Randomized Trees. Our methods ranked 10th in the first track and 4th in the second track. Although not at the very top of the ranking, our results show that ensembles of randomized trees are quite competitive for the “learning to rank” problem. The paper also analyzes computing times of our algorithms and presents some post-challenge experiments with transfer learning methods. [less ▲]

Detailed reference viewed: 335 (73 ULg)
Peer Reviewed
See detailLearning Visual Representations for Interactive Systems
Piater, Justus ULg; Jodogne, Sébastien ULg; Detry, Renaud ULg et al

in 14th International Symposium on Robotics Research (2009)

Detailed reference viewed: 9 (4 ULg)
Full Text
Peer Reviewed
See detailLearning Visual Representations for Perception-Action Systems
Piater, Justus ULg; JODOGNE, Sébastien ULg; Detry, Renaud ULg et al

in International Journal of Robotics Research (2011), 30(3), 294-307

Detailed reference viewed: 11 (6 ULg)
Peer Reviewed
See detailLearning, then Compacting Visual Policies
Jodogne, Sébastien ULg; Piater, Justus ULg

in 7th European Workshop on Reinforcement Learning (2005)

Detailed reference viewed: 4 (3 ULg)