HAL - Lille 3
Not a member yet
5967 research outputs found
Sort by
DIOGEN, a multi-level oriented model for cartographic generalization
No full textInternational audienceAmong approaches for automated generalization of vector data, we focus on the multi-agent paradigm: cartographic objects are modeled as agents (autonomous objects) that apply generalization algorithms to themselves to satisfy cartographic constraints. Several agent levels are considered, for example, individual agents, such as a building, and agents representing a group of agents, such as an urban block composed of the surrounding roads and contained buildings. Several multi-agent models were proposed to automate the orchestration of map generalization processes. Existing multi-agent generalization models have different approaches to manage the relations between agent levels. In this paper, we unify existing models, adapting a multi-level simulation model, to simplify interactions between agents in different levels. We propose the DIOGEN model, in which the principle of interactions between agents of different levels is adapted to constraint-driven cartographic generalization. DIOGEN unifies three existing multi-agent generalization models (AGENT, CartACom and GAEL), combine their behaviors and take advantage of their skills. Our proposal is evaluated on different use cases: instances of topographic mapping, and mapping of hiking routes over topographic data as an example of thematic mapping.Nous nous intéressons aux approches dédiées à la généralisation automatique basées sur le paradigme multi-agents: les objets cartographiques sont modélisés comme des agents (objets autonomes) qui s'appliquent des algorithmes de généralisation pour satisfaire des contraintes cartographiques. Plusieurs niveaux d'agents sont considérés, par exemple des agents individuels, comme un bâtiment, et des agents représentant un groupe d'agents, comme un îlot urbain composé des routes qui l'entourent et des bâtiments qu'il contient. Plusieurs modèles multi-agents ont été proposés pour automatiser l'orchestration d'un processus de généralisation. Les modèles existants gèrent différemment les relations entre les niveaux d'agents. Dans cet article, nous travaillons sur l'unification des modèles existants. Nous simplifions les interactions entre agents des différents niveaux en adaptant un modèle agent récemment défini pour la simulation, et qui met l'accent sur la modélisation du multi-niveau. La modélisation des interactions entre niveaux issue de ce modèle est adaptée au cas de la généralisation cartographique guidée par des contraintes. Le modèle résultant s'appelle DIOGEN. Il unifie trois modèles de génralisation existants (AGENT, CartACom et GAEL), permettant de combiner leurs comportements et leurs capacités. Notre proposition est évaluée sur des cas concrets de cartographie topographique, ainsi que sur de la cartographie conjointe d'itinéraires de randonnée et de données topographiques qui constitue un exemple de cartographie thématique
Decentralized Collaborative Learning of Personalized Models over Networks
No full textInternational audienceWe consider a set of learning agents in a col-laborative peer-to-peer network, where each agent learns a personalized model according to its own learning objective. The question addressed in this paper is: how can agents improve upon their locally trained model by communicating with other agents that have similar objectives? We introduce and analyze two asynchronous gossip algorithms running in a fully decentralized manner. Our first approach , inspired from label propagation, aims to smooth pre-trained local models over the network while accounting for the confidence that each agent has in its initial model. In our second approach, agents jointly learn and propagate their model by making iterative updates based on both their local dataset and the behavior of their neighbors. To optimize this challenging objective, our decentralized algorithm is based on ADMM
Delexicalized Word Embeddings for Cross-lingual Dependency Parsing
No full textInternational audienceThis paper presents a new approach to the problem of cross-lingual dependency parsing, aiming at leveraging training data from different source languages to learn a parser in a target language. Specifically , this approach first constructs word vector representations that exploit structural (i.e., dependency-based) contexts but only considering the morpho-syntactic information associated with each word and its contexts. These delexicalized word em-beddings, which can be trained on any set of languages and capture features shared across languages, are then used in combination with standard language-specific features to train a lexicalized parser in the target language. We evaluate our approach through experiments on a set of eight different languages that are part the Universal Dependencies Project. Our main results show that using such delexicalized embeddings, either trained in a monolin-gual or multilingual fashion, achieves significant improvements over monolingual baselines
Online Learning of Task-specific Word Representations with a Joint Biconvex Passive-Aggressive Algorithm
No full textInternational audienceThis paper presents a new, efficient method for learning task-specific word vectors using a variant of the Passive-Aggressive algorithm. Specifically, this algorithm learns a word embedding matrix in tandem with the classifier parameters in an online fashion, solving a bi-convex constrained optimization at each iteration. We provide a theoretical analysis of this new algorithm in terms of regret bounds, and evaluate it on both synthetic data and NLP classification problems, including text classification and sentiment analysis. In the latter case, we compare various pre-trained word vectors to initialize our word embedding matrix, and show that the matrix learned by our algorithm vastly outperforms the initial matrix, with performance results comparable or above the state-of-the-art on these tasks
Parameter Exploration to Improve Performance of Memristor-Based Neuromorphic Architectures
No full textInternational audienceThe brain-inspired spiking neural network neuromorphic architecture offers a promising solution for a wide set of cognitive computation tasks at a very low power consumption. Due to the practical feasibility of hardware implementation, we present a memristor-based model of hardware spiking neural networks which we simulate with N2S3 (Neural Network Scalable Spiking Simulator), our open source neuromorphic architecture simulator. Although Spiking neural networks are widely used in the community of computational neuroscience and neuromorphic computation, there is still a need for research on the methods to choose the optimum parameters for better recognition efficiency. With the help of our simulator, we analyze and evaluate the impact of different parameters such as number of neurons, STDP window, neuron threshold, distribution of input spikes and memristor model parameters on the MNIST handwritten digit recognition problem. We show that a careful choice of a few parameters (number of neurons, kind of synapse, STDP window and neuron threshold) can significantly improve the recognition rate on this benchmark (around 15 points of improvement for the number of neurons, a few points for the others) with a variability of 4 to 5 points of recognition rate due to the random initialization of the synaptic weights
Le nouvel élan des sciences humaines et sociales : « Le vase qui parle »
No full textInternational audienceVidéo de l'ntervention présentant le projet du Vase qui parle lors de la 2e édition du forum des nouvelles initiatives en médiation scientifique. Durée de l'intervention : 14 min
Baudelaire : El "jovial mistificador" de la prensa literaria y artistica
No full textInternational audienc
Baudelaire: traductor-auctoritas
No full textInternational audienceIn order to achieve visibility in the media and a position recognized by both the public and their peers, translators are compelled to take advantage of spaces of enunciation such as those provided by prefaces, criticism, or biographical notes. Thanks to these spaces, in which translators deploy discursive and institutional strategies that allow them to position themselves and their translation project, translators acquire the status of translator-aucforitas, that is, a level of symbolic authority capable of endowing them with a public image. Through the detailed analysis of the editorial strategies and institutional calculations implemented by Baudelaire in order to position his project of translating Edgar Allan Poe, we show how the poet achieves the status of translator-auctoritas and the role the latter played in the construction of his own literary identity
Caregivers' quality of life and psychological health in response to functional, cognitive, neuropsychiatric and social deficits of patients with brain tumour: protocol for a cross-sectional study
No full textINTRODUCTION: Patients with gliomas generally present cognitive, neuropsychiatric and functional deficits. Although previous research has shown that their caregivers present a poor quality of life and poor mental health, only a few studies have tested in a comprehensive way which deficits/preserved abilities of patients predominantly impact their caregivers. Furthermore, only a few studies have focused on the social impact of gliomas, which may also damage the caregivers' quality of life. Therefore, this cross-sectional study aims to investigate which patients' impairments are particularly deleterious for the caregivers and whether the histological characteristics of the gliomas also affect their quality of life.METHODS AND ANALYSIS: In order to examine these research questions, this study intends to include 180 patients (60 patients with grade II gliomas, 60 patients with grade III gliomas and 60 patients with grade IV gliomas), their caregivers and 60 healthy controls. While patients will complete a full battery of cognitive, neuropsychiatric, functional and social tests, caregivers will complete questionnaires about their quality of life, depression, anxiety and burden. Patients' performances and caregivers' reports of depression and anxiety will be compared with the scores of healthy controls. Eventually, our aim will be to provide specific care support both to reduce patients' deficits and alleviate caregivers' difficulties.ETHICS AND DISSEMINATION: The study has obtained the approval of the local faculty ethics committee ('Comité d'éthique en sciences comportementales'; 2016-5 S41 and 2015-3 S37). On completion of the study, data will be kept by Lille University for 5 years before they are destroyed. Study findings will be disseminated through peer-reviewed journal publications and conference presentations with no reference to a specific individual