BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//LORIA - ECPv6.17.3.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://www.loria.fr
X-WR-CALDESC:Évènements pour LORIA
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20230326T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20231029T010000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20240331T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20241027T010000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20250330T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20251026T010000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20240315T100000
DTEND;TZID=Europe/Paris:20240315T110000
DTSTAMP:20240313T131459Z
CREATED:20240313T131422Z
LAST-MODIFIED:20240313T131459Z
UID:25126-1710496800-1710500400@www.loria.fr
SUMMARY:D1 Seminar: Compute-Based Rendering across the Board: Efficient Methods for Meshes\, Point Clouds and Radiance Fields
DESCRIPTION:Next D1 Seminar\, entitled « Compute-Based Rendering across the Board: Efficient Methods for Meshes\, Point Clouds and Radiance Fields »\, will take place on March 15 at 10 am\, in room A006. \nAbstract:  \nDr. Kerbl will present his on-going research on compute-based rendering and its use across 3D representations and applications. Modern 3D content\, both synthetic and captured\, contains an unprecedented amount of geometric detail. At the same time\, user appreciation of 3D content demands low-latency feedback loops: processing\, exploring or modifying 3D scenes should be fast\, or—even better—instant. Failure to meet performance targets is often answered by applying more raw compute power to the problem. This policy has led to systematic hardware hoarding\, a rift between « GPU-rich » and « GPU-poor »\, increased reliance on cloud computing and an overall rise in global resource consumption. A more sustainable solution to this challenge lies in the careful design of inherently parallel algorithms\, finding novel data structures and optimal 3D scene representations for specific tasks. One pillar of the research by Dr. Kerbl et al. towards this goal is the use of compute-based rendering: exploiting GPU compute to assist or replace the fixed-function pipeline for image formation. This talk illustrates concrete examples where compute-based rendering achieves or surpasses state-of-the-art results\, given only a fraction of its competitors’ runtime resources. Apart from image formation itself\, this talk will also discuss recent applications in interactive editing and differentiable rendering methods for radiance fields (NeRFshop\, 3D Gaussian Splatting). \nBio:  \nAfter a Bachelor and Master’s degree in information technology\, Dr. Kerbl received his PhD from Graz University of Technology in 2018 for his research on GPU workload scheduling. In 2019\, he briefly joined Epic Games to work on the Nanite Virtual Geometry feature of Unreal Engine 5. This was followed by a postdoc research stay at TU Wien and another at Inria\, Université Côte d’Azur in the GraphDeco group. Currently\, he acts as principal investigator on a project acquired in 2022 for Instant Visualization and Interaction for Point Clouds (IVILPC). Dr. Kerbl has taught multiple courses on the design and programming of modern GPU hardware at three Austrian universities and has (co-)supervised several students from undergraduate to PhD level.
URL:https://www.loria.fr/event/d1-seminar-compute-based-rendering-across-the-board-efficient-methods-for-meshes-point-clouds-and-radiance-fields/
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20240329T100000
DTEND;TZID=Europe/Paris:20240329T110000
DTSTAMP:20240321T084412Z
CREATED:20240321T084412Z
LAST-MODIFIED:20240321T084412Z
UID:25415-1711706400-1711710000@www.loria.fr
SUMMARY:Machine Learning at the Service of Decomposition Techniques in Mixed Integer Programming: A cross docking assignment and scheduling problem
DESCRIPTION:Shahin Gelareh\, Maître de conférences à l’université d’Artois\, présentera ses travaux de recherche au laboratoire le vendredi 29 mars à 10h en salle C005. \nTitle : Machine Learning at the Service of Decomposition Techniques in Mixed Integer Programming: A cross docking assignment and scheduling problem\nAbstract: The primary objective of this presentation is to share our experiences and illustrate how two distinct classes of machine learning algorithms can be seamlessly integrated into decomposition frameworks to expedite the solution process’s convergence. We exemplify this through the truck dock assignment and scheduling problem\, an operational issue that requires frequent resolution throughout the day whenever the existing plan is disrupted by unforeseen events. The operational nature of this problem is crucial as the data distribution remains relatively stable over a considerable period\, facilitating the accumulation of ample training data without issues caused by distribution shifts. Our focus is predominantly on two exact methods: Cut-and-Benders and Dantzig-Wolfe. However\, given enough time\, we also demonstrate that the same trained deep learning model can assist in constructing feasible solutions and can be incorporated into a reinforcement learning model to function as a very efficient heuristic. \nRésumé : L’objectif principal de cette présentation est de partager nos expériences et d’illustrer comment deux classes distinctes d’algorithmes d’apprentissage automatique peuvent être intégrées dans des cadres de décomposition pour accélérer la convergence du processus de résolution. Nous exemplifions ceci à travers le problème d’attribution et de planification de quais pour camions\, un problème opérationnel nécessitant une résolution fréquente au cours de la journée\, chaque fois que le plan existant est interrompu par des événements imprévus. La nature opérationnelle de ce problème est essentielle car la distribution des données reste relativement stable sur une période considérable\, facilitant ainsi l’accumulation d’une quantité suffisante de données d’entraînement sans les problèmes causés par les changements de distribution. Notre attention se porte principalement sur deux méthodes exactes : Cut-and-Benders et Dantzig-Wolfe. Cependant\, si le temps nous permet\, nous démontrons également que le même modèle d’apprentissage profond entraîné peut aider à construire des solutions réalisables et peut être intégré dans un modèle d’apprentissage par renforcement pour fonctionner comme une heuristique très efficace.
URL:https://www.loria.fr/event/machine-learning-at-the-service-of-decomposition-techniques-in-mixed-integer-programming-a-cross-docking-assignment-and-scheduling-problem/
LOCATION:C005
CATEGORIES:Séminaire
END:VEVENT
END:VCALENDAR