Centre de documentation
scientifique

Towards expressive graph neural networks : Theory, algorithms, and applications / Georgios Dasoulas (2022)  

Public Titre : Towards expressive graph neural networks : Theory, algorithms, and applications Type de document : Thèse/HDR Auteurs : Georgios Dasoulas, Auteur ; Michalis Vazirgiannis, Directeur de thèse ; Aladin Virmaux, Directeur de thèse Editeur : Paris : Institut Polytechnique de Paris Année de publication : 2022 Note générale : bibliographie These de doctorat de l’Institut Polytechnique de Paris préparée à l’Ecole Polytechnique, spécialité Informatique Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Intelligence artificielle [Termes IGN] apprentissage automatique [Termes IGN] attention (apprentissage automatique) [Termes IGN] entropie [Termes IGN] graphe [Termes IGN] isomorphisme [Termes IGN] noeud [Termes IGN] réseau neuronal de graphes [Termes IGN] théorie des graphes Index. décimale : THESE Thèses et HDR Résumé : (auteur) As the technological evolution of machine learning is accelerating nowadays, data plays a vital role in building intelligent models, being able to simulate phenomena, predict values and make decisions. In an increasing number of applications, data take the form of networks. The inherent graph structure of network data motivated the evolution of the graph representation learning field. Its scope includes generating meaningful representations for graphs and their components, i.e., the nodes and the edges. The research on graph representation learning was accelerated with the success of message passing frameworks applied on graphs, namely the Graph Neural Networks. Learning informative and expressive representations on graphs plays a critical role in a wide range of real-world applications, from telecommunication and social networks, urban design, chemistry, and biology. In this thesis, we study various aspects from which Graph Neural Networks can be more expressive, and we propose novel approaches to improve their performance in standard graph learning tasks. The main branches of the present thesis include: the universality of graph representations, the increase of the receptive field of graph neural networks, the design of stable deeper graph learning models, and alternatives to the standard message-passing framework. Performing both theoretical and experimental studies, we show how the proposed approaches can become valuable and efficient tools for designing more powerful graph learning models.In the first part of the thesis, we study the quality of graph representations as a function of their discrimination power, i.e., how easily we can differentiate graphs that are not isomorphic. Firstly, we show that standard message-passing schemes are not universal due to the inability of simple aggregators to separate nodes with ambiguities (similar attribute vectors and neighborhood structures). Based on the found limitations, we propose a simple coloring scheme that can provide universal representations with theoretical guarantees and experimental validations of the performance superiority. Secondly, moving beyond the standard message-passing paradigm, we propose an approach for treating a corpus of graphs as a whole instead of examining graph pairs. To do so, we learn a soft permutation matrix for each graph, and we project all graphs in a common vector space, achieving a solid performance on graph classification tasks.In the second part of the thesis, our primary focus is concentrated around the receptive field of the graph neural networks, i.e., how much information a node has in order to update its representation. To begin with, we study the spectral properties of operators that encode adjacency information. We propose a novel parametric family of operators that can adapt throughout training and provide a flexible framework for data-dependent neighborhood representations. We show that the incorporation of this approach has a substantial impact on both node classification and graph classification tasks. Next, we study how considering the k-hop neighborhood information for a node representation can output more powerful graph neural network models. The resulted models are proven capable of identifying structural properties, such as connectivity and triangle-freeness.In the third part of the thesis, we address the problem of long-range interactions, where nodes that lie in distant parts of the graph can affect each other. In this problem, we either need the design of deeper models or the reformulation of how proximity is defined in the graph. Firstly, we study the design of deeper attention models, focusing on graph attention. We calibrate the gradient flow of the model by introducing a novel normalization that enforces Lipschitz continuity. Next, we propose a data augmentation method for enriching the node attributes with information that encloses structural information based on local entropy measures. Note de contenu : 1. Introduction 2. Preliminaries I- Discrimination power 3. Universal approximation on graphs 4. Learning soft permutations for graph representations II- Receptive field 5. Learning graph shift operators 6. Increasing the receptive field with multiple hops III- Beyond local interactions 7. Lipschitz continuity of graph attention 8. Structural symmetries in graphs 9. Conclusions and outlook Numéro de notice : 24076 Affiliation des auteurs : non IGN Thématique : INFORMATIQUE/MATHEMATIQUE Nature : Thèse française Note de thèse : Thèse de doctorat : Informatique : Palaiseau : 2022 DOI : sans En ligne : https://theses.hal.science/tel-03666690 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=102200

A polyhedra-based model for moving regions in databases / Florian Heinz in International journal of geographical information science IJGIS, vol 34 n° 1 (January 2020)  

Public [article]  inInternational journal of geographical information science IJGIS > vol 34 n° 1 (January 2020) . - pp 41 - 73 Titre : A polyhedra-based model for moving regions in databases Type de document : Article/Communication Auteurs : Florian Heinz, Auteur ; Ralf Hartmut Güting, Auteur Année de publication : 2020 Article en page(s) : pp 41 - 73 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Systèmes d'information géographique [Termes IGN] base de données orientée objet [Termes IGN] CGAL [Termes IGN] implémentation (informatique) [Termes IGN] isomorphisme [Termes IGN] MADS [Termes IGN] modélisation spatio-temporelle [Termes IGN] objet mobile [Termes IGN] polyèdre Résumé : (auteur) Moving objects databases store and process objects with a focus on their spatiotemporal behaviour. To achieve this, the model of the data must be suitable to efficiently store and process moving objects. Currently, a unit-based model is widely used, where each moving object is divided into one or more time intervals, during which the object behaves uniformly. This model is also used for a data type called moving regions, which resembles moving and shape changing regions as, for example, forest fires or cloud fields. However, this model struggles to support operations like union, difference or intersection of two moving regions; the resulting objects are unnecessarily bloated and uncomfortable to handle because the resulting number of units is generally very high. In this paper, an alternative model for moving regions is proposed, which is based on polyhedra. Furthermore, this work develops an isomorphism between moving regions and polyhedra including all relevant operations, which has the additional advantage that several implementations for those are already readily available; this is demonstrated by a reference implementation using the existing and well-tested Computational Geometry Algorithms Library (CGAL). Numéro de notice : A2020-007 Affiliation des auteurs : non IGN Thématique : GEOMATIQUE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1080/13658816.2019.1616090 Date de publication en ligne : 17/05/2019 En ligne : https://doi.org/10.1080/13658816.2019.1616090 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=94387 [article]

Exemplaires(1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
079-2020011	RAB	Revue	Centre de documentation	En réserve L003	Disponible

The “M” in digital elevation models / XiaoHang Liu in Cartography and Geographic Information Science, Vol 42 n° 3 (July 2015)  

Public [article]  inCartography and Geographic Information Science > Vol 42 n° 3 (July 2015) . - pp 235 - 243 Titre : The “M” in digital elevation models Type de document : Article/Communication Auteurs : XiaoHang Liu, Auteur ; Hai Hu, Auteur ; Peng Hu, Auteur Année de publication : 2015 Article en page(s) : pp 235 - 243 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications photogrammétriques [Termes IGN] données lidar [Termes IGN] généralisation sémantique [Termes IGN] isomorphisme [Termes IGN] modèle numérique de surface [Termes IGN] modélisation 3D Résumé : (auteur) The “M” in digital elevation models (DEM) stands for model, which literally means “a schematic description of a system, theory, or phenomenon that accounts for its known or inferred properties and may be used for further study of its characteristics.” A DEM fulfills the requirement of “a schematic description” of terrain. However, how to make it account for the “known or inferred properties” warrants further scrutiny. This article outlines three properties of terrain and examines their four implications to DEM generation. The three properties are as follows: (1) each terrain point has a single, fixed elevation; (2) terrain points have an order and sequence that is determined by their elevations; and (3) terrain has skeletons. The four implications to DEM generation methods are as follows: (1) a method must be a bijection; (2) a method must be an isomorphism in order to preserve elevation sequence; (3) a method must guarantee that the vertical error at any point, not just checkpoints, is acceptable in order to assure the vertical accuracy of a DEM; and (4) a method must involve generalization if terrain skeletons are to be preserved. These implications are discussed in the context of light detection and ranging-derived DEMs. Generalization is highlighted as the top priority for future research. Numéro de notice : A2015-241 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article DOI : 10.1080/15230406.2014.993711 En ligne : https://doi.org/10.1080/15230406.2014.993711 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=76239 [article]

Exemplaires(1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
032-2015031	RAB	Revue	Centre de documentation	En réserve L003	Disponible

Mise en correspondance et reconstruction stéréo utilisant une description structurelle des images / Thomas Skordas (1988)  

Public Titre : Mise en correspondance et reconstruction stéréo utilisant une description structurelle des images Type de document : Thèse/HDR Auteurs : Thomas Skordas, Auteur ; Radu Horaud, Directeur de thèse Editeur : Grenoble : Institut National Polytechnique de Grenoble INPG Année de publication : 1988 Format : 21 x 30 cm Note générale : bibliographie Thèse présentée pour obtenir le titre de Docteur de l'Institut National Polytechnique de Grenoble, Spécialité Informatique Langues : Français (fre) Descripteur : [Vedettes matières IGN] Traitement d'image [Termes IGN] appariement d'images [Termes IGN] contrainte géométrique [Termes IGN] extraction de traits caractéristiques [Termes IGN] graphe relationnel [Termes IGN] isomorphisme [Termes IGN] reconstruction 3D [Termes IGN] reconstruction d'objet [Termes IGN] scène [Termes IGN] vision monoculaire [Termes IGN] vision par ordinateur [Termes IGN] vision stéréoscopique Index. décimale : THESE Thèses et HDR Résumé : (auteur) Description d'un système de vision stéréoscopique. Le but est de mettre en correspondance deux images d'une même scène acquises de deux points de vue différents et de reconstituer ensuite la structure tridimensionnelle de la scène. Des segments de droite sont extraits de l'image de gauche et de l'image de droite. Ces segments sont groupés convenablement afin de former des structures locales 2d. Sur la base de ces groupements, chaque image est décrite par un graphe relationnel ou les nœuds représentent les segments et les arcs représentent des relations structurelles entre les segments. La mise en correspondance des deux descriptions monoculaires obtenues est réalisée en trouvant des structures similaires entre les deux descriptions, ce qui est équivalent à la recherche d'un isomorphisme. Note de contenu : Introduction 1- Description monoculaire d'une image 2- La mise en correspondance 3- La reconstruction de surface Conclusions générales Numéro de notice : 21745 Affiliation des auteurs : non IGN Thématique : IMAGERIE/INFORMATIQUE Nature : Thèse française Note de thèse : Thèse de Doctorat : Informatique : INPG Grenoble : 1988 nature-HAL : Thèse DOI : sans En ligne : https://tel.archives-ouvertes.fr/tel-00329721/document Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=91078

Cartographie polonaise / Poland Head Office Of Geodesy And Cartography (1987)

Public Titre : Cartographie polonaise Type de document : Monographie Auteurs : Poland Head Office Of Geodesy And Cartography, Auteur Editeur : Varsovie : Head Office of Geodesy and Cartography Année de publication : 1987 Importance : 169 p. Format : 17 x 24 cm ISBN/ISSN/EAN : 978-83-7000-061-5 Langues : Anglais (eng) Français (fre) Espagnol (spa) Descripteur : [Vedettes matières IGN] Cartographie étrangère [Termes IGN] Antarctique [Termes IGN] canevas [Termes IGN] carte hydrographique [Termes IGN] cartographie automatique [Termes IGN] cartographie thématique [Termes IGN] cartographie topographique [Termes IGN] classification [Termes IGN] Cracovie (Pologne) [Termes IGN] géodésie [Termes IGN] historique [Termes IGN] homomorphisme [Termes IGN] impression cartographique [Termes IGN] impression offset [Termes IGN] isomorphisme [Termes IGN] parc naturel national [Termes IGN] paysage [Termes IGN] Pologne Numéro de notice : 57494 Affiliation des auteurs : non IGN Thématique : GEOMATIQUE Nature : Monographie Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=60218

Exemplaires(1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
57494-01	39.40	Livre	Centre de documentation	Cartographie	Disponible

An alternative isomorphism for mapping correlation / Mark Monmonier in Annuaire international de cartographie, n° 19 (1979)

Permalink

Eléments de mathématique. Fascicule 23, Livre 2 Algèbre, chapitre 8 Modules et anneaux semi-simples / Nicolas Bourbaki (1973)  

Permalink

Leçons sur les projections des cartes géographiques / Albert Reyt (1967)

Permalink

Eléments de mathématiques. Fascicule 22, Théorie des ensembles, chapître 4 Structures / Nicolas Bourbaki (1966)  

Permalink

Leçons sur les projections des cartes géographiques / Albert Reyt (1960)

Permalink