BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//LORIA - ECPv6.17.4.1//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
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20260329T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20261025T010000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250918T130000
DTEND;TZID=Europe/Paris:20250918T150000
DTSTAMP:20250829T144242Z
CREATED:20250829T144242Z
LAST-MODIFIED:20250829T144242Z
UID:28347-1758200400-1758207600@www.loria.fr
SUMMARY:Séminaire Sécurité Loria (SSL) : Ksenia Ermoshina & Charlie Jacomme
DESCRIPTION:Next SSL (Loria Security Seminar) will take place on Thursday\, September 18th at 1pm in room A008. \nFor this come back after the summer\, we will have an “hors format” edition with two talks: \n\nKsenia Ermoshina\, CNRS researcher at Centre Internet et Société.\nDecentralized e2ee messaging apps: a challenge for security and usability?\nCharlie Jacomme\, Inria researcher in the Pesto team (Loria)\nSecurity Analysis of the Signal Messenger specifications and implementation\n\n\nKsenia Ermoshina (Centre Internet et Société).\nDecentralized e2ee messaging apps: a challenge for security and usability? \n\nThis presentation explores the ecosystem of federated end-to-end encrypted messaging applications\, and the challenges that decentralized architectures raise in terms of security and usability. Based on an in-depth ethnographic study of the communities involved in developing\, testing and using federated messengers (namely Element and Delta Chat)\, this research explores the unprecedented success of decentralized platforms and protocols in Europe and across the world. Indeed\, in the recent years\, the trust of users in centralized and proprietary platforms and messengers have been eroding\, especially among marginalized communities\, activists\, journalists and tech enthusiasts. Following the acquisition of X by Elon Musk and the recent inauguration of Trump in January 2025\, the question of digital sovereignty and the urgency of migrating from US-owned messengers have become central. We have analyzed how various communities\, including governments in France and Germany\, have adopted Matrix protocol for their communications. In this presentation we will address the technological and sociopolitical issues related to the adoption of federated apps by the civil society and public institutions. \n\n\n\nCharlie Jacomme (Inria Nancy\, Loria):\nSecurity Analysis of the Signal Messenger specifications and implementation\nSignal Messenger is one of the most widely used secure messaging application\, and with its underlying library also used by WhatsApp and Facebook Messenger\, it provides end-to-end encryption to millions of users everyday.  In this talk\, we will present the protocols underlying the Signal Messenger application\, which mainly consists of three distinct specifications: PQXDH for the initial key exchange\, the Double Ratchet for updating encryption keys for each message sent out\, and Sesame\, for managing  multiple devices and sessions. While introducing each of the components and their articulation\, we will present how for several years we have been using a mix of formal methods and real world experiments to study the security guarantees provided by the application. We will present several positive results of security obtained with computer-aided cryptography\, as well as describe several attacks we uncovered\, some only on the design and the specifications of the protocol\, and some others having consequences on the actual forward secrecy or post-compromise security of the implementation. We will finally describe how we interacted with Signal’s developers to propose fixes\, most of which were integrated in the specifications or the implementation.
URL:https://www.loria.fr/event/seminaire-securite-loria-ssl-ksenia-ermoshina-charlie-jacomme/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250917T140000
DTEND;TZID=Europe/Paris:20250917T160000
DTSTAMP:20250829T154230Z
CREATED:20250829T154230Z
LAST-MODIFIED:20250829T154230Z
UID:28349-1758117600-1758124800@www.loria.fr
SUMMARY:DeepLorIA Tutorial: Franco Terranova (Resist)
DESCRIPTION:DeepLorIA is an internal network for sharing knowledge about artificial intelligence\, whatever the field of application. It is open to all Loria’s teams that use AI in their research. \nNext DeepLorIA tutorial will take place on September 17th from 2 to 4 pm in room A008\, with Franco Terranova (Resist team). \nMeta-Reinforcement Learning: Boosting RL Generalization \nReinforcement Learning (RL) enables agents to learn optimal behaviors through trial and error but often struggles to generalize to new situations. Meta-Reinforcement Learning (Meta-RL) addresses this limitation by leveraging prior experience to help agents adapt more quickly to novel tasks. This tutorial will cover the fundamentals of deep RL and meta-learning\, including an introduction to MAML applied to RL. The hands-on session will use a grid environment to explore how policies trained on a single environment\, with domain randomization\, or via meta-learning respond to environmental perturbations. \nSchedule: Theory 14:00 – 14:50 / Coffee Break 14:50 – 15:10 / Lab 15:10 – 16:00 \n? Register here
URL:https://www.loria.fr/event/deeploria-tutorial-franco-terranova-resist/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250708T110000
DTEND;TZID=Europe/Paris:20250708T120000
DTSTAMP:20250617T113521Z
CREATED:20250617T113521Z
LAST-MODIFIED:20250617T113521Z
UID:28250-1751972400-1751976000@www.loria.fr
SUMMARY:MosAIK team seminar: Willem Waegeman (Ghent University)
DESCRIPTION:The MosAIK team organizes a seminar open to all the laboratory on Tuesday\, 8 July at 11am in room A008. \nWillem Waegeman\, associate professor at Ghent University\, will give a presentation entitled Making AI systems more trustworthy through uncertainty disentanglement. \nAbstract:\nGiven the increasing use of machine learning (ML) models for decisions that directly affect humans\, it is essential that these models not only provide accurate predictions but also offer a credible representation of their uncertainty. Recent advances have led to probabilistic models capable of disentangling two types of uncertainty: aleatoric and epistemic. Aleatoric uncertainty is inherent to the data and cannot be eliminated\, while epistemic uncertainty is related to the ML model and can be reduced with better modeling approaches or more data. In this talk I will elaborate on the opportunities and limitations of uncertainty disentanglement in explaining why an ML model fails to deliver accurate predictions. Furthermore\, I will discuss several use cases that demonstrate the potential of uncertainty disentanglement for biotechnology applications. \nBio:\n \nWillem Waegeman is an associate professor at Ghent University\, and group leader of the BIOML group of the Department of Data Analysis and Mathematical Modelling. His main research interests are machine learning and bioinformatics. Specific interests include uncertainty quantification and complex prediction problems\, such as multi-target and structured prediction problems. He is an author of more than 100 peer-reviewed papers in journals and conferences\, and his work has won several prizes. In recent years he has served on the program committees of leading conferences in AI (ICML\, Neurips\, ECML/PKDD\, ICLR\, UAI\, AISTATS\, IJCAI\, etc.).
URL:https://www.loria.fr/event/mosaik-team-seminar-willem-waegeman-ghent-university/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250620T094500
DTEND;TZID=Europe/Paris:20250620T120000
DTSTAMP:20250620T072623Z
CREATED:20250620T072616Z
LAST-MODIFIED:20250620T072623Z
UID:28260-1750412700-1750420800@www.loria.fr
SUMMARY:Matinée de présentation des nouveaux permanents du D4
DESCRIPTION:Le Département 4 du LORIA organise une demi-journée de présentation de ses nouveaux permanents le vendredi 20 juin de 9 h 45 à 12 h en salle A008. \nProgramme : \n9 h 45 : accueil café\n10 h : François Buet (MCF Télécom Nancy\, MosAIk)\n10 h 30 : Chahrazed Labba (MCF INSPE\, MosAIk)\n11 h : Vincent Martin (CR Inria\, Sémagramme)\n11 h 30 : discussions.
URL:https://www.loria.fr/event/matinee-de-presentation-des-nouveaux-permanents-du-d4/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250617T140000
DTEND;TZID=Europe/Paris:20250617T160000
DTSTAMP:20250522T130915Z
CREATED:20250522T130747Z
LAST-MODIFIED:20250522T130915Z
UID:28222-1750168800-1750176000@www.loria.fr
SUMMARY:[DeepLorIA] Mining Patterns in Tabular Data – A Symbolic AI Approach
DESCRIPTION:DeepLorIA est un réseau interne pour partager des connaissances en intelligence artificielle\, quel que soit le domaine d’application. Il est ouvert à toutes les équipes du Loria qui utilisent l’IA dans leurs travaux. \nLe prochain tutoriel DeepLorIA aura lieu le 17 juin\, de 14h à 16h\, en salle C005. Il est proposé par Egor Dudyrev (Orpailleur). \nMining Patterns in Tabular Data – A Symbolic AI Approach\nThis session will cover Symbolic AI\, which is a vast research field operating human-readable knowledge representations. In particular\, we will focus on Pattern Structures formalism that provides a unified language to efficiently mine patterns in tables containing numbers\, categories\, ngrams\, graphs\, and other types of data. Such patterns can then be used for mining explainable-by-design implications\, clusterings\, and multiclass classifiers. The hand-on-section will explore a novel Paspailleur Python package (developed in Loria\, set to release in May 2025) that can solve all aforementioned tasks for all aforementioned data. \nInscrivez-vous ! \nPlus d’informations auprès de Franco Terranova\, doctorant dans l’équipe Resist.
URL:https://www.loria.fr/event/deeploria-mining-patterns-in-tabular-data-a-symbolic-ai-approach/
LOCATION:C005
CATEGORIES:Séminaire
ATTACH;FMTTYPE=image/png:https://www.loria.fr/wp-content/uploads/DeepLorIA_logo1.png
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250610T130000
DTEND;TZID=Europe/Paris:20250610T143000
DTSTAMP:20250603T140426Z
CREATED:20250603T140400Z
LAST-MODIFIED:20250603T140426Z
UID:28244-1749560400-1749565800@www.loria.fr
SUMMARY:SSL: Loria Security Seminar - Guillaume Scerri (ENS Paris-Saclay\, LMF)
DESCRIPTION:Next Loria Security Seminar will take place on Tuesday\, 10th June at 1pm in room B011-B013. \nGuillaume Scerri (ENS Paris-Saclay\, LMF)\, will give a presentation entitled Capturing new cryptographic proofs techniques using logic: extension to the computationally complete symbolic attacker. \nWhen proving cryptographic protocols\, one has to deal with a malicious attacker. In particular this means that it is crucial to carefully capture attacker capabilities. This is best done by reducing security of protocols to known hard problems. However these cryptographic reductions can be quite complex and when done by hand such reductions can be rather hard to check. In recent years there has been a push to capture such reductions using logics that can be checked using (dedicated) proof assistants. In this talk we explore the intricacies of capturing complex cryptographic reductions in one of these logics the Computationally Complete Symbolic Attacker logic. We will focus on two main techniques: hybrid arguments and rewinding\, and show how they can be applied to proofs of new protocols\, namely e-voting mixnets.
URL:https://www.loria.fr/event/ssl-loria-security-seminar/
LOCATION:B011-B013
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250610T090000
DTEND;TZID=Europe/Paris:20250610T173000
DTSTAMP:20250506T074632Z
CREATED:20250506T072700Z
LAST-MODIFIED:20250506T074632Z
UID:28111-1749546000-1749576600@www.loria.fr
SUMMARY:Journée scientifique du Pôle AM2i et de la Chambre d'Agriculture
DESCRIPTION:Le pôle AM2I organise une journée scientifique commune avec la Chambre d’Agriculture le  mardi 10 juin 2025 à la Chambre d’Agriculture (Laxou) de 9 heures à 17 h 30.\n\nCette journée est l’occasion d’échanger sur les compétences et les questions de recherche en agriculture numérique\, ce qui inclut la robotique\, l’IA\, la science des données\, le traitement du signal (capteurs\, image\, ….)\, la modélisation statistique\, l’optimisation\,  ainsi que de nombreux autres thèmes présents dans nos laboratoires.\n\nAu programme :\n\nAccueil-café à partir de 9h\, début à 9h30 :\n\n– présentation des activités de la Chambre d’Agriculture\, du pôle AM2I\, et des possibilités de coopération\n– présentation de 3 projets en partenariat avec la Chambre d’Agriculture (robotique\, big data\, modélisation statistique)\n\n– buffet et temps d’échange.\n\nà partir de 14h :\n\n– présentation de projets de recherche menées dans la pôle AM2I (robotique\, modélisation\, image\, traitement de données)\n– table ronde « Utilisation de l’IA dans le Conseil Agricole »\n\nFin à 17h30\n\n\nInscription (gratuite mais obligatoire) : avant le vendredi 30 mai sur ce lien.\n\nAccès : Chambre d’Agriculture\, 9 rue de la Vologne\, Laxou (Bus T2 arrêt Vologne).\n\nComité d’organisation : Anne Incerti\, Angelo Koudou\, Antoine Lejay\, Frédéric Pierlot\, Thibaut Raharijaona
URL:https://www.loria.fr/event/journee-scientifique-du-pole-am2i-et-de-la-chambre-dagriculture/
LOCATION:Chambre d’Agriculture\, Laxou
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250514T140000
DTEND;TZID=Europe/Paris:20250514T160000
DTSTAMP:20250415T124754Z
CREATED:20250415T124716Z
LAST-MODIFIED:20250415T124754Z
UID:27957-1747231200-1747238400@www.loria.fr
SUMMARY:[DeepLorIA] Graph Neural Networks – A Hands-On Guide to Graph Data Processing
DESCRIPTION:DeepLorIA est un réseau interne pour partager des connaissances en intelligence artificielle\, quel que soit le domaine d’application. Il est ouvert à toutes les équipes du Loria qui utilisent l’IA dans leurs travaux. \nLe prochain tutoriel DeepLorIA aura lieu le 14 mai\, de 14h à 16h\, en salle A008. Il est proposé par Victor Pryakhin (Capsid). \nGraph Neural Networks – A Hands-On Guide to Graph Data Processing\nThis session introduces Graph Neural Networks (GNNs) and their innovative approach to understanding complex graphs. Dive into the core concepts—from graph representations and message passing to pooling and cutting-edge attention mechanisms. Using PyTorch Geometric\, you’ll gain practical experience building\, training\, and optimizing GNN models on real-world datasets spanning molecular structures\, social networks\, and beyond. \nInscrivez-vous ! \nPlus d’informations auprès de Franco Terranova\, doctorant dans l’équipe Resist.
URL:https://www.loria.fr/event/deeploria-graph-neural-networks/
LOCATION:A008
CATEGORIES:Séminaire
ATTACH;FMTTYPE=image/png:https://www.loria.fr/wp-content/uploads/DeepLorIA_logo1.png
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250428T100000
DTEND;TZID=Europe/Paris:20250428T110000
DTSTAMP:20250331T162258Z
CREATED:20250331T162258Z
LAST-MODIFIED:20250331T162258Z
UID:27882-1745834400-1745838000@www.loria.fr
SUMMARY:[NLP seminar] Behaving according to protocol: How communicative projects are carried out differently in different contexts
DESCRIPTION:The NLP Seminar « Behaving according to protocol: How communicative projects are carried out differently in different contexts »\, by Ellen Breitholtz\, will take place on Monday\, 28th April\, at 10 a.m.\, in Loria’s room A008. \nAbstract: There are a number of theories and models for capturing the aspects of organisations that are systematically related to the modes and genres of communication taking place within them. In this talk I will consider the micro-level of organisations and communities and present a model of how similar communicative projects are carried out differently within different activities. Central to this account is the notion of conversational games\, which can be seen as strategies for realising communicative projects while assigning speaker roles to dialogue participants. \nBio: Ellen Breitholtz is lecturer at Gothenburg University. Her research focuses on formal modeling of dialogue and on various aspects of pragmatic reasoning. She works on Enthymemes (logically incomplete argument) and Topoi (principles of reasoning) in Dialogue. She also participate in different research project on NLP and mental health.
URL:https://www.loria.fr/event/nlp-seminar-behaving-according-to-protocol-how-communicative-projects-are-carried-out-differently-in-different-contexts/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250415T130000
DTEND;TZID=Europe/Paris:20250415T143000
DTSTAMP:20250410T090923Z
CREATED:20250410T090923Z
LAST-MODIFIED:20250410T090923Z
UID:27915-1744722000-1744727400@www.loria.fr
SUMMARY:Loria Security Seminar - André Schrottenloher (IRISA)
DESCRIPTION:Next Loria Security Seminar will take place on Tuesday\, 15 April 2025 at 1pm at Loria. \nAndré Schrottenloher (Univ Rennes\, Inria\, CNRS\, IRISA) will give a presentation entitled « Reducing the Number of Qubits in Quantum Factoring« . \nShor’s algorithm (FOCS 1994) is arguably the most powerful application of quantum algorithms to cryptanalysis\, as it solves the factoring and discrete logarithm problem in polynomial time. In the past 30 years\, many authors have optimized its time complexity (the number of basic operations\, also known as quantum gates) and memory footprint (the number of qubits). While quantum computers are still far from running Shor’s algorithm in practice\, these fine-grained optimizations improve our understanding of the resources required to do so. \nIn this talk\, I will present a new memory optimization in this algorithm. As in previous works\, we focus on the modular exponentiation modulo N\, which is its core component. We show that a logarithmic number of work qubits suffices to obtain the least significant bits of the output. We combine this result with May and Schlieper’s truncation technique (ToSC 2022) and the Ekera-Hastad variant of Shor’s algorithm (PQCrypto 2017) to solve the discrete logarithm problem in ℤN* using only d + o(log N) qubits\, where d is the bit-size of the logarithm. Consequently we can factor n-bit RSA moduli using n/2 + o(n) qubits\, a quarter of the previous most optimized implementations.
URL:https://www.loria.fr/event/loria-security-seminar-andre-schrottenloher-irisa/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250411T110000
DTEND;TZID=Europe/Paris:20250411T120000
DTSTAMP:20250326T162129Z
CREATED:20250326T162129Z
LAST-MODIFIED:20250326T162129Z
UID:27858-1744369200-1744372800@www.loria.fr
SUMMARY:[Séminaire ABC] Isotropic Kernel Machine
DESCRIPTION:Le prochain séminaire de l’équipe ABC\, intitulé Isotropic Kernel Machine\, aura lieu le 11 avril à 11h\, en salle Alice & Bob.\nL’exposant\, Nicolas Wicker\, est Professeur de mathématiques à l’Université de Lille et membre du laboratoire Paul Painlevé.\nAbstract:\nA new kernel machine for multi-class pattern recognition is introduced: the isotropic kernel machine. It is designed to make use of the isotropy of the class conditional densities in the feature space. We provide theoretical guarantees on its generalization error. This error is then assessed empirically\, in the framework of a comparative study.
URL:https://www.loria.fr/event/seminaire-abc-isotropic-kernel-machine/
LOCATION:Loria
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;VALUE=DATE:20250407
DTEND;VALUE=DATE:20250409
DTSTAMP:20250219T095356Z
CREATED:20250219T094947Z
LAST-MODIFIED:20250219T095356Z
UID:27575-1743984000-1744156799@www.loria.fr
SUMMARY:CAIPI : Endomorphismes et invariants de variétés abéliennes
DESCRIPTION:CAIPI est un séminaire de mathématiques sur la théorie des Codes correcteurs\, la cryptogrAphie\, la géométrie arIthmétique\, et leurs aspects comPutationnels\, dans un format Itinérant. \nLa prochaine édition\, « Endomorphismes et invariants de variétés abéliennes »\, aura lieu les 7 et 8 avril au Loria. Trois exposés sont proposés par : \n\nAnnamaria Iezzi\nLaboratoire Jean Kuntzmann\, Université Grenoble Alpes\nElisa Lorenzo García\nUniversité de Neuchâtel\, Université de Rennes\nStefano Marseglia\nLaboratoire J.A. Dieudonné\, Université Côte d’Azur\n\nPlus d’informations sur le site de l’événement. \nContact : caipi_symposium-request[at]services.cnrs.fr
URL:https://www.loria.fr/event/caipi-endomorphismes-et-invariants-de-varietes-abeliennes/
LOCATION:Loria
CATEGORIES:Séminaire
ATTACH;FMTTYPE=image/png:https://www.loria.fr/wp-content/uploads/CAIPI-symposium.png
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250404T140000
DTEND;TZID=Europe/Paris:20250404T150000
DTSTAMP:20250331T161822Z
CREATED:20250331T161821Z
LAST-MODIFIED:20250331T161822Z
UID:27880-1743775200-1743778800@www.loria.fr
SUMMARY:Unveiling Mixture of Experts for Multi-Sensor Fusion and Multi-Task Learning
DESCRIPTION:The seminar « Unveiling Mixture of Experts for Multi-Sensor Fusion and Multi-Task Learning« \, by Zongwei Wu\, will take place on Friday\, April 4\, at 2 p.m.\, in room C103\, at the Loria. \nAbstract: Building toward a foundation visual model for real-world perception requires processing diverse sensor inputs while simultaneously handling multiple tasks. Achieving this demands efficient yet effective knowledge transfer across different domains—whether sensors or tasks. While recent Mixture of Experts (MoE) architectures\, such as Mistral and DeepSeek\, have demonstrated remarkable efficiency in large language models\, they are not explicitly designed for cross-domain knowledge sharing.\nIn this talk\, I will present our work on two novel MoE variations tailored for multi-sensor fusion and multi-task reasoning\, respectively. For multi-sensor fusion\, we introduce a Mixture of Modal Experts (MeME)\, a model that dynamically selects and integrates modality-specific experts to enhance cross-modal learning\, improving robustness in uncertain environments. More importantly\, MeME enables emergent alignment using only paired data.\nFor multi-task learning\, we propose a Mixture of Complexity Experts (MoCE)\, an architecture that adapts expert selection based on task complexity. This design naturally transforms MoCE into a task-discriminative Learner\, facilitating efficient multi-task learning with adaptive resource allocation.\nThrough these innovations\, we show how MoE can go beyond efficiency\, enabling effective cross-domain knowledge sharing for the next generation of perception models. \nBio: Zongwei Wu is a PostDoc Researcher at the Computer Vision Lab\, University of Würzburg\, Germany. He received his diplôme d’ingénieur from the University of Technology of Compiègne in 2019 and earned a Ph.D. in Computer Vision from Vibot EMR CNRS 6000\, University of Burgundy\, France in 2022. He was also a visiting scholar at CVL\, ETH Zurich. His research focuses on multimodal models and multi-task reasoning for machine vision. He is a main organizer of the NTIRE workshop at CVPR 2024-2025 and serves as an Associate Editor for IEEE RA-L.
URL:https://www.loria.fr/event/unveiling-mixture-of-experts-for-multi-sensor-fusion-and-multi-task-learning/
LOCATION:Loria\, salle C103
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250402T140000
DTEND;TZID=Europe/Paris:20250402T150000
DTSTAMP:20250321T152819Z
CREATED:20250321T152819Z
LAST-MODIFIED:20250321T152819Z
UID:27811-1743602400-1743606000@www.loria.fr
SUMMARY:Apprentissage continu et détection d’anomalies : vers des systèmes fiables en environnements ouverts
DESCRIPTION:Le séminaire « Apprentissage continu et détection d’anomalies : vers des systèmes fiables en environnements ouverts »\, proposé par Son Vu\, Maître de conférences à l’ENSEA et chercheur au laboratoire ETIS\, aura lieu au Loria\, le 2 avril à 14h\, en salle C103. \nRésumé: Bien que les modèles profonds aient connu un succès remarquable dans diverses applications\, ils ne sont pas sans limites. Un inconvénient majeur réside dans leur capacité à exceller principalement dans des environnements fermés\, tout en peinant à s’adapter à des situations imprévues ou à des événements rares. Cette fragilité peut devenir problématique dans des systèmes critiques comme l’imagerie médicale\, où des cas atypiques peuvent fausser un diagnostic\, ou la formation chirurgicale en réalité virtuelle\, où une erreur peut induire une mauvaise décision.\nDans cette présentation\, nous partagerons quelques récentes avancées pour autonomiser les systèmes afin qu’ils fonctionnent de manière sûre et autonome dans des scénarios ouverts\, avec un focus sur la détection d’anomalies et l’apprentissage continu. En ce qui concerne la détection d’anomalies\, nous proposons le premier modèle unifié AnoMem combinant mémoire Hopfield et apprentissage auto-supervisé\, adapté aux scénarios “one-class (zero-shot)” et “unbalanced outlier-exposure (few-shot)”. Nous présenterons également SeeABLE et CTru\, deux algorithmes pionniers proposant des solutions innovantes pour la détection de deepfakes\, en reformulant cette tâche comme un problème de détection d’anomalies zero-shot dans des espaces hypersphériques et hyperboliques. Ensuite\, nous dévoilerons DSDM\, un nouvel algorithme d’apprentissage continu exploitant les réseaux de mémoire associative dynamique.\nEnfin\, nous discuterons de la manière dont l’apprentissage continu et la détection d’anomalies peuvent bénéficier à d’autres tâches\, telles que le recalage et la localisation visuelle.
URL:https://www.loria.fr/event/apprentissage-continu-et-detection-danomalies-vers-des-systemes-fiables-en-environnements-ouverts/
LOCATION:Loria\, salle C103
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250401T130000
DTEND;TZID=Europe/Paris:20250401T143000
DTSTAMP:20250314T142746Z
CREATED:20250314T142746Z
LAST-MODIFIED:20250314T142746Z
UID:27805-1743512400-1743517800@www.loria.fr
SUMMARY:Department 2 Seminar: Kostia Chardonnet
DESCRIPTION:The next formal methods seminar will take place on April 1st at 13:00 in room A008. \nSpeaker: Kostia Chardonnet \nTitle: Infinite Algebraic Proofs \nLors du séminaire du département du 7 janvier\, Alejandro a montré comment étendre le fragment multiplicatif-additif de la logique linéaire\nintuitionniste au cas quantique. En particulier\, il a montré comment représenter\, à travers un système de preuve et par Curry-Howard\, un langage de programmation pour toute application linéaire entre des espaces de Hilbert de dimension finie. \nDans cet exposé\, nous nous intéressons à la représentation similaire\, mais pour des espaces de dimension infinie. Pour ce faire\, nous nous appuyons sur les méthodes provenant des logiques à points fixes.
URL:https://www.loria.fr/event/department-2-seminar-kostia-chardonnet/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250327T140000
DTEND;TZID=Europe/Paris:20250327T150000
DTSTAMP:20250325T143135Z
CREATED:20250325T142927Z
LAST-MODIFIED:20250325T143135Z
UID:27842-1743084000-1743087600@www.loria.fr
SUMMARY:[NLP Seminar] What does Emotion have to do with Language?
DESCRIPTION:Next NLP Seminar will take place on Thursday\, 27th March\, at 2 p.m.\, in room A008 Jean Legras. \nSpeaker:\nCasey Kennington\, Boise State University \nAbstract:\nResearch suggests that cognition is not separable from emotion\, which has implications for computational models of language: they are trained on text (and in some cases images)\, but how is emotion represented? In my talk\, we explore 4E cognition\, where emotion fits\, how we can model emotion\, and how emotion and language can come together in computational models. \nBio:\nOn the occasion of the Master TAL graduation ceremony\, Casey Kennington agreed to be the class sponsor. Casey is an alumnus of the Master’s program who is now an associate professor in the Department of Computer Science at Boise State University. He is head of the Speech\, Language and Interactive Machines group at Boise State University and active in the SIGDIAL community. He principally does research in interactive spoken dialogue systems\, semantics\, human-robot interaction\, and language acquisition.
URL:https://www.loria.fr/event/nlp-seminar-what-does-emotion-have-to-do-with-language/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250310T100000
DTEND;TZID=Europe/Paris:20250310T110000
DTSTAMP:20250224T094441Z
CREATED:20250224T094441Z
LAST-MODIFIED:20250224T094441Z
UID:27594-1741600800-1741604400@www.loria.fr
SUMMARY:NLP Seminar: Watermarking AI-generated content
DESCRIPTION:The next NLP Seminar\, entitled « Watermarking AI-generated content » will be given by Pierre Fernandez (Meta FAIR Paris) and will take place on 10th March\, in Loria’s room A008 Jean Legras\, at 10:00. \nAbstract:\nInvisible watermarking involves embedding a hidden signal into digital media like text\, images\, and audio. It has evolved from a tool for copyright protection to a critical solution for AI safety and security\, for example to enhance the detection of AI-generated content. For Generative AI\, watermarking can be implemented at various stages: post-hoc or at generation-time. We will explore this across modalities\, relying on the following works done in the lab: \n\nProactive Detection of Voice Cloning with Localized Watermarking https://arxiv.org/abs/2401.17264\nThree Bricks to Consolidate Watermarks for Large Language Models https://arxiv.org/abs/2308.00113\nThe Stable Signature: Rooting Watermarks in Latent Diffusion Models https://arxiv.org/abs/2303.15435\nLatent Watermarking of Audio Generative Models https://arxiv.org/abs/2409.02915\n\nThe presentation will conclude with a discussion on the limitations and future prospects of invisible watermarking. \nBio:\nPierre Fernandez is a research scientist at Meta FAIR Paris (Meta Fundamental AI Research lab). His research focuses on watermarking\, particularly for generative models to ensure AI-generated content is traceable and identifiable. During his PhD at FAIR and Inria Rennes\, he notably developed the audio watermarking algorithm used in FAIR’s public research demos like Audiobox and Seamless\, as well as the image watermarking algorithm for Meta AI image generation assistant (Imagine).
URL:https://www.loria.fr/event/nlp-seminar-watermarking-ai-generated-content/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250307T090000
DTEND;TZID=Europe/Paris:20250307T100000
DTSTAMP:20250304T172756Z
CREATED:20250304T170353Z
LAST-MODIFIED:20250304T172756Z
UID:27697-1741338000-1741341600@www.loria.fr
SUMMARY:D3 Seminar : Analyse temporelle de systèmes temps réel complexes avec le Network Calculus
DESCRIPTION:Le prochain séminaire « Analyse temporelle de systèmes temps réel complexes avec le Network Calculus » sera présenté par Damien Guidolin-Pina (ISAE SUPAERO/ONERA/RTaW)\, le 7 mars à 9h en salle A008. \nRésumé\nLes systèmes temps réel sont aujourd’hui de plus en plus complexes et différentes technologies se mêlent. Afin de garantir que le système respecte les exigences temporelles (latences\, gigues\, etc.) sans surdimensionner\, il est nécessaire d’analyser le plus précisément possible un système. Pour ce faire\, différentes théories existent et permettent de calculer des bornes assez précises. Cependant\, elles ne le permettent que sur certaines parties du système (des bornes locales). \nPour calculer une borne globale\, une théorie est souvent soit pessimiste soit pas développée pour analyser toutes les parties d’un système. Ainsi\, il est commun d’utiliser plusieurs théories pour calculer les meilleures bornes locales et les sommer pour obtenir une borne globale. Cependant\, le peu d’information échangée entre les théories engendre du pessimisme dans le calcul des bornes globales. L’objectif ici est d’améliorer l’évaluation des bornes de délai globale (de bout en bout) des systèmes temps réel critiques. Deux approches sont étudiées : \n1. Améliorer la compatibilité du Network Calculus pour le lier avec par exemple Compositional Performance Analysis\, afin d’exploiter leurs forces respectives sans introduire de pessimisme excessif dû au manque de partage d’informations. Le défi majeur réside dans la gestion des différences de continuité des courbes utilisées dans ces théories\, et un mécanisme de conversion est proposé pour assurer une compatibilité formelle. \n2. Étendre le domaine d’application du Network Calculus pour modéliser directement des éléments typiques des systèmes temps réel\, notamment les tâches\, les\nchaînes de tâches déclenchées et les hyperviseurs. Une formalisation de l’exécution des tâches est introduite\, intégrant des concepts de mise à l’échelle et de packetization pour capturer la propagation des dépendances temporelles. L’analyse des hyperviseurs est faite en tenant compte de la synchronisation des partitions et de l’alignement des flux avec leurs fenêtres d’exécution. Une nouvelle approche permet de mieux intégrer ces contraintes temporelles dans le cadre du Network Calculus\, offrant ainsi une modélisation plus fidèle des architectures temps réel complexes. \nBio :\nDamien Guidolin-Pina a obtenu le diplôme d’ingénieur de l’ISAE ENSMA\, à Poitiers\, France\, en 2020. Il a ensuite rejoint RealTime-at-Work\, un éditeur de logiciels français aidant les OEMs et Tier1s à concevoir des architectures de communication électriques/électroniques (E/E) sûres et optimisées en termes de coûts.\nPendant dix mois\, en tant qu’ingénieur logiciel\, il a travaillé sur l’implémentation de fonctionnalités notamment autour de IEEE TSN dans RTaW-PEGASE\, une plateforme leader de l’industrie pour la conception\, la configuration et la simulation d’architectures de communication embarquées. Il a ensuite réalisé son doctorat chez RealTime-at-Work\, de 2021 à 2024\, au sein de l’ONERA Toulouse. Son directeur de thèse était Marc Boyer (ONERA/DTIS). Ses recherches portent sur l’analyse temporelle des systèmes de communication temps réel\, principalement en utilisant la théorie du calcul des réseaux et la configuration de mécanisme IEEE TSN.
URL:https://www.loria.fr/event/d3-seminar-analyse-temporelle-de-systemes-temps-reel-complexes-avec-le-network-calculus/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250305T100000
DTEND;TZID=Europe/Paris:20250305T110000
DTSTAMP:20250226T085205Z
CREATED:20250226T085114Z
LAST-MODIFIED:20250226T085205Z
UID:27671-1741168800-1741172400@www.loria.fr
SUMMARY:[Séminaire DefMal] La recherche en cybersécurité doit-elle s’intéresser davantage à la criminologie (et vice-versa) ?
DESCRIPTION:Dans le cadre du projet cible du PEPR Cybersécurité – DefMal porté par l’Université de Lorraine\, le séminaire « La recherche en cybersécurité doit-elle s’intéresser davantage à la criminologie (et vice-versa) ?« \, sera présenté par Benoît Dupont\, professeur titulaire à l’École de criminologie de l’Université de Montréal et titulaire de la Chaire de recherche du Canada en Cyber-résilience ainsi que de la Chaire de recherche en Prévention de la cybercriminalité. \nL’événement aura lieu au Loria le mercredi 5 mars à 10h\, en salle A008. \nRésumé\nLa cybercriminalité désigne à la fois les délits traditionnels facilités par les technologies numériques (crimes cyber-facilités) et les infractions qui n’existeraient pas sans elles (crimes cyber-dépendants). De son côté\, la cybersécurité englobe les technologies\, politiques et pratiques visant à protéger les données\, réseaux et systèmes contre les intrusions. Ces deux champs disciplinaires\, bien que distincts\, sont de plus en plus liés. La cybercriminalité est perçue comme une menace majeure pour la cybersécurité\, incitant les institutions de sécurité à s’impliquer davantage dans sa prévention et son contrôle. Ce séminaire soulignera la nécessité d’un rapprochement entre la cybercriminologie et la cybersécurité\, encore trop cloisonnées. En nous appuyant sur un examen détaillé des cadres théoriques et des méthodes empiriques des deux domaines\, je propose de considérer cybercriminalité et cybersécurité comme appartenant au même continuum. Cette approche favorise une meilleure compréhension des liens entre ces deux domaines et encourage des collaborations plus étroites entre chercheurs en cybercriminalité et en cybersécurité. \n\nLes personnes externes au laboratoire souhaitant assister à l’événement doivent contacter Maira Nassau (prénom.nom@loria.fr) avant lundi 3 mars.
URL:https://www.loria.fr/event/seminaire-defmal-la-recherche-en-cybersecurite-doit-elle-sinteresser-davantage-a-la-criminologie-et-vice-versa/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250304T130000
DTEND;TZID=Europe/Paris:20250304T140000
DTSTAMP:20250122T135802Z
CREATED:20250122T135651Z
LAST-MODIFIED:20250122T135802Z
UID:27111-1741093200-1741096800@www.loria.fr
SUMMARY:[Formal Methods Seminar] A scalable framework for backward bounded static symbolic execution
DESCRIPTION:The next D2 seminar\, entitled “A scalable framework for backward bounded static symbolic execution”\, by Nicolas Bellec\, will be held on March 4 at 1:00 pm in room A008. \nAbstract\nMany programs (e.g. malware) hide their behavior by using obfuscations such as opaque predicates. Automatic methods have been developed to detect such obfuscations. In this presentation\, we will focus on static symbolic backward bounded execution\, a method that enumerates backward bounded paths from a potential opaque predicate and uses symbolic execution to check whether all enumerated paths have the same predicate value. \nWe’ll show how this method can be generalized using a lattice-based framework and how to scale up such methods to efficiently analyze whole programs. \nMore information on the department’s website
URL:https://www.loria.fr/event/formal-methods-seminar-3/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250228T103000
DTEND;TZID=Europe/Paris:20250228T120000
DTSTAMP:20250204T090445Z
CREATED:20250131T132628Z
LAST-MODIFIED:20250204T090445Z
UID:27256-1740738600-1740744000@www.loria.fr
SUMMARY:D5 Seminar: Interactive molecular visualization and simulation
DESCRIPTION:The next D5 Seminar\, « Interactive molecular visualization and simulation: examples with UDock and VTX » will be held by Pr. Matthieu Montes\, from Le CNAM\, laboratory of Génomique\, bioinformatique et chimie moléculaire (GBCM). The event will take place on Friday\, February 28 at 10:30 a.m. in room A008. \nAbstract:\nUDock (available at https://udock.fr and on gitlab): Originally developed to gamify protein-protein docking\, UDock is now an interactive multibody molecular docking software designed to explore protein interactions for research purposes but also for teaching and popularization of science due to its high usability. In UDock\, the users tackle the conformational space of protein interfaces using an intuitive real-time docking procedure with on-the-fly scoring. UDock2 integrates traditional computer graphics methods to facilitate the visualization and to provide better insight into protein surfaces\, interfaces\, and properties. \n VTX (available at http://vtx.drugdesign.fr and on github)\, is a high-performance molecular visualization software. VTX is optimized to handle efficiently the big data from High Performance Computing Molecular Dynamics Simulation. Its high-performance 3D engine includes cutting-edge computer graphics methods capable of handling large and dynamic molecular scenes on a standard laptop computer. It also offers a video game based minimalistic task-oriented graphical user interface to maximize its usability and comfort of use.
URL:https://www.loria.fr/event/d5-seminar-interactive-molecular-visualization-and-simulation/
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250204T130000
DTEND;TZID=Europe/Paris:20250204T140000
DTSTAMP:20250122T135849Z
CREATED:20250122T135322Z
LAST-MODIFIED:20250122T135849Z
UID:27108-1738674000-1738677600@www.loria.fr
SUMMARY:[Formal Methods Seminar] The Squirrel Prover
DESCRIPTION:The next D2 seminar\, entitled “The Squirrel Prover”\, by Charlie Jacomme\, will be held on February 4 at 1:00 pm in room A008. \nAbstract\nThe Squirrel Prover is a proof assistant dedicated to cryptographic protocols. It relies on a higher-order logic following the computationally complete symbolic attacker approach. It thus provides guarantees in the computational model. In this talk\, we will introduce the main ingredients underlying its logic and proof system\, trying to outline why it does yield computational guarantees and how it allows to reason over protocols. We will then cover the current development status of Squirrel\, showcasing its current features and expressivity. \nMore information on the department’s website
URL:https://www.loria.fr/event/formal-methods-seminar-the-squirrel-prover/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250121T130000
DTEND;TZID=Europe/Paris:20250121T140000
DTSTAMP:20250110T112405Z
CREATED:20250110T112405Z
LAST-MODIFIED:20250110T112405Z
UID:27063-1737464400-1737468000@www.loria.fr
SUMMARY:SSL - Loria Security Seminar - Nils Ole Tippenhauer (CISPA)
DESCRIPTION:Next SSL – Loria Security Seminar will take place on Tuesday\, 21st January at 1pm in room A008. \nNils Ole Tippenhauer\, researcher at CISPA\, will talk about reversing and testing cyber-physical systems.  \nTitle: Reversing and Testing Cyber-Physical Systems \nAbstract: Cyber-Physical Systems such as Industrial Control Systems\, Drones\, and Autonomous Vehicles rely on digital control of a physical process. Adversarial manipulation of such systems can result in physical damage to the system\, or the environment. The underlying embedded systems run (often proprietary) firmware on bare metal\, interacting with a range of peripherals. Security assessment of such systems by their operators\, vendors\, and attackers is challenging for various reasons. In this talk\, I will go over some of those challenges\, and discuss recent ongoing work in the direction of reversing\, rehosting\, and fuzzing of such systems. \nBio: Nils is a faculty at the CISPA Helmholtz Center for Information Security\, heading the SCy-Phy research group. Prior to joining CISPA in 2018\, he was an Assistant Professor at the Singapore University of Technology and Design (SUTD). Nils earned his Dr. Sc. in Computer Science from ETH Zurich (Switzerland) in 2012.  His research is broadly on information security aspects of practical systems. In particular\, he is am currently working on security of cyber-physical systems such as drones\, vehicles\, and industrial control systems. \nAhh the information about the seminar is on the SSL website.
URL:https://www.loria.fr/event/ssl-loria-security-seminar-nils-ole-tippenhauer-cispa/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20250107T130000
DTEND;TZID=Europe/Paris:20250107T150000
DTSTAMP:20241211T134814Z
CREATED:20241211T134814Z
LAST-MODIFIED:20241211T134814Z
UID:27006-1736254800-1736262000@www.loria.fr
SUMMARY:Department 2 seminar: Alejandro Díaz-Caro (Mocqua team)
DESCRIPTION:Next Formal Methods Seminar (D2 seminar) will take place on January 7th at 1pm in room A008. \nAlejandro Díaz Caro (Mocqua) Will give a presentation entitled Sup\, Sum\, and Scalars: In Linear Logic and in Non-Linear Logic. \nAbstract:\nIn [1]\, we introduced a new connective\, called « sup »\, to intuitionistic\npropositional logic to model information erasure\, non-reversibility\, and\nnon-determinism\, as observed in quantum measurement\, among other\ncontexts. This connective features the introduction rule of conjunction\nand the elimination rule of disjunction\, resulting in a\nnon-deterministic cut-elimination process. To enrich its proof terms\, we\nadded a sum operator\, which allows for recovering the determinism of the\nconnective. Additionally\, since one of our goals was to express quantum\nprograms\, we introduced a scalar product. These rules may occupy the\nspace between the introduction and elimination of a given connective\,\nforcing us to include commuting cuts to establish a proper introduction\ntheorem.\nThe next step toward transforming this calculus into a quantum calculus\nwas to adopt a linear logic framework. Thus\, in [2]\, we considered the\nfull intuitionistic multiplicative-additive linear logic (IMALL). We\nproved that\, when removing the sup connective\, the system–augmented\nwith sums and scalar multiplication–is linear in the algebraic sense.\nSpecifically\, any term t applied to a linear combination a⋅r+b⋅s is\nobservationally equivalent to a⋅(tr)+b⋅(ts).\nIn [3]\, we provided an abstract characterization of the sup connective\nin IMALL\, showing that any symmetric monoidal closed category with\nbiproducts and a monomorphism from the semiring of scalars to the\nsemiring Hom(I\,I) is a suitable model. By leveraging the binary\nbiproducts\, we defined a weighted codiagonal map\, which serves as the\ncore of the sup connective.\nFinally\, in [4]\, we considered two non-linear systems: first\, with the\nsum operator alone as a parallel operator\, and second\, with the\ncombination of the sum and scalar operators as an algebraic lambda\ncalculus. We proposed two categorical models for these calculi. In\nparticular\, we explored the role of disjunction within the calculus\,\nwhich has rarely been studied in conjunction with the parallel (or sum)\nconstruct. For the parallel lambda calculus\, we employed the category\nMagSet\, whose objects are magmas and whose morphisms are functions from\nthe category Set. For the algebraic lambda calculus\, we used the\ncategory AMagSSet\, whose objects are action magmas\, with morphisms\nsimilarly drawn from Set. Our approach diverges from conventional\ninterpretations\, which typically handle disjunctions via coproducts.\nInstead\, we proposed handling them through a combination of disjoint\nunions and Cartesian products.
URL:https://www.loria.fr/event/department-2-seminar-alejandro-diaz-caro-mocqua-team/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20241213T133000
DTEND;TZID=Europe/Paris:20241213T150000
DTSTAMP:20241209T151832Z
CREATED:20241209T151832Z
LAST-MODIFIED:20241209T151832Z
UID:26999-1734096600-1734102000@www.loria.fr
SUMMARY:D5 Seminar: Bio-inspired self-supervised learning of visual representations
DESCRIPTION:The next D5 Seminar\, « Bio-inspired self-supervised learning of visual representations » will be held by Arthur Aubret\, on Friday\, December 13 at 1:30 p.m. in room C005. \nAbstract:\nAbstract: Despite recent advances in self-supervised visual machine learning\, humans develop more robust representations with much fewer data. This may be explained by the fundamental differences between the development of their visual systems: while machine learning methods use massive amounts of i.i.d images\, humans actively move and interact with objects over time. In this talk\, I investigate how considering bio-inspired learning mechanisms can impact visual representations learning. I will provide evidence that modelling spatio-temporal regularities in egocentric visual sequences boosts the robustness of vision models. In addition\, I will explain how egocentric actions underpinning visual changes\, like eye saccades or object manipulations\, support object learning. Together\, these findings expose that the spatio-temporal structure and active nature of human visual experience may be key to develop strong semantic visual representations.
URL:https://www.loria.fr/event/d5-seminar-bio-inspired-self-supervised-learning-of-visual-representations/
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20241206T133000
DTEND;TZID=Europe/Paris:20241206T150000
DTSTAMP:20241129T133542Z
CREATED:20241129T133542Z
LAST-MODIFIED:20241129T133542Z
UID:26976-1733491800-1733497200@www.loria.fr
SUMMARY:D5 Seminar: Deep Learning\, Optimal Control\, and Bio-Inspired Control for Dynamic Robots
DESCRIPTION:The next D5 Seminar\, « Deep Learning\, Optimal Control\, and Bio-Inspired Control for Dynamic Robots » will be held by Guillaume Bellegarda (EPFL) on Friday\, December 6 at 1:30 p.m. in room A008. \nAbstract:\nRecent advances in machine learning\, control\, and robotics show promising results towards integrating autonomous systems into society. Legged robots in particular suggest potential for the same dynamic capabilities as humans to adapt to everyday\, and even challenging\, environments. However\, when compared with humans and animals\, state-of-the-art robotic systems do not yet demonstrate the same agility nor intelligence to navigate the real world. While important contributions have been made to approach human levels in specific tasks\, the desired system generalizability to interact with and adapt to new environments remains challenging. Additionally\, for tasks learned with machine learning in which robotic systems do approach or surmount human-level skills\, the underlying neural network function approximation lacks interpretability and performance guarantees. This is true for both Artificial Neural Networks\, as well as their biological counterparts that exist in animals. In this talk\, we present several methods to maximize robotic system performance and explainability by leveraging ideas from machine learning\, model-based control\, and neuroscience. Example applications will be shown for highly dynamic motions on systems such as quadrupeds\, vehicles\, and wheel-legged robots.
URL:https://www.loria.fr/event/d5-seminar-deep-learning-optimal-control-and-bio-inspired-control-for-dynamic-robots/
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20241203T130000
DTEND;TZID=Europe/Paris:20241203T140000
DTSTAMP:20241115T093532Z
CREATED:20241115T093532Z
LAST-MODIFIED:20241115T093532Z
UID:26921-1733230800-1733234400@www.loria.fr
SUMMARY:Department 2 Seminar: Thibault Gauthier.
DESCRIPTION:Le prochain séminaire du D2 aura lieu le 3 décembre à 13h00 en salle A008. \nOrateur : Thibault Gauthier \nTitre : The Automation of Mathematics in Practice \nAbstract:\nThis presentation provides an overview of automated systems actively\nused in proving mathematical theorems\, focusing on the practical\nsuccesses  achieved by SAT solvers\, first-order theorem provers\, and\ninteractive theorem provers.\nAdditionally\, we present our progress towards improving automation in\nareas that are usually considered challenging for proof automation\, such\nas conjecture generation and inductive reasoning. \nToutes les infos se trouvent aussi sur le site du département.
URL:https://www.loria.fr/event/department-2-seminar-thibault-gauthier/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;VALUE=DATE:20241127
DTEND;VALUE=DATE:20241128
DTSTAMP:20241106T123605Z
CREATED:20241106T123605Z
LAST-MODIFIED:20241106T123605Z
UID:26878-1732665600-1732751999@www.loria.fr
SUMMARY:Journée du Pôle AM2I "Logistique 4.0 & Supply Chain intelligente et durable"
DESCRIPTION:La première journée scientifique du Pôle AM2I\, sur le thème Logistique 4.0 & Supply Chain intelligente et durable : Optimisation \, Robotique\, Résilience et Digitalisation\, aura lieu le  mercredi 27 novembre 2024 dans les nouveaux locaux de l’Institut d’Innovation en Logistique au Technopôle de Metz.\n\nElle abordera différents aspects de la recherche en logistique ainsi que les travaux réalisés dans nos laboratoires\, et les défis à surmonter en automatique\, robotique\, analyse de données (big data\, analytics)\, IoT\, IA\, réalité augmentée ou virtuelle\, …\n\nCette journée se compose de 4 sessions thématiques\, d’une conférence et une table ronde :\n\n Conférence – Les tendances actuelles en logistique et supply chain.\n    Session 1 – Logistique pour les énergies  (LGIPM\, I2L).\n    Session 2 – Résilience et Optimisation (LORIA\, LGIPM)\n    Session 3 – Robotique et Logistique (LCFC\, LCOMS/LGIPM\, société TECH3D).\n    Session 4 – Logistique durable et digitalisation de la supply chain (CEREFIGE).\n\nTable ronde – Entrepreneuriat et innovation exploration pour la logistique et la supply chain.\n\nInscription sur ce formulaire.
URL:https://www.loria.fr/event/journee-du-pole-am2i-logistique-4-0-supply-chain-intelligente-et-durable/
LOCATION:Institut d’Innovation en Logistique\, Tecnopôle de Metz.\, Avenue de Strasbourg - Technopole 2\, Metz\, 57000\, France
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20241119T130000
DTEND;TZID=Europe/Paris:20241119T143000
DTSTAMP:20241021T140657Z
CREATED:20241008T125425Z
LAST-MODIFIED:20241021T140657Z
UID:26733-1732021200-1732026600@www.loria.fr
SUMMARY:SSL - Loria Security Seminar - Geoffroy Couteau
DESCRIPTION:The Loria Security Seminar (SSL – séminaire sécurité du Loria) is back! \nThis seminar aims to offer a comprehensive overview of security topics to all interested people (students\, postdoc\, permanent members\, engineers\, etc). \nOur first seminar will be given by Geoffroy Couteau\, CNRS researcher at IRIF\, on November 19th\, 2024 at 1:00pm in room A008.\nAll Inria or Loria members are welcome.\n\n\nTitle: Fast Public-Key Silent OT and More from Constrained Naor-Reingold\n\nAbstract: Pseudorandom Correlation Functions (PCFs) allow two parties\, given correlated evaluation keys\, to locally generate arbitrarily many pseudorandom correlated strings\, e.g. Oblivious Transfer (OT) correlations\, which can then be used by the two parties to jointly run secure computation protocols.\nIn this work\, we provide a novel and simple approach for constructing PCFs for OT correlation\, by relying on constrained pseudorandom functions for a class of constraints containing a weak pseudorandom function (wPRF). We then show that tweaking the Naor-Reingold pseudorandom function and relying on low-complexity pseudorandom functions allow us to instantiate our paradigm. We further extend our ideas to obtain efficient public-key PCFs\, which allow the distribution of correlated keys between parties to be non-interactive: each party can generate a pair of public/secret keys\, and any pair of parties can locally derive their correlated evaluation key by combining their secret key with the other party’s public key. \nIn addition to these theoretical contributions\, we detail various optimizations and provide concrete instantiations of our paradigm relying on the Boneh-Ishai-Passelègue-Sahai-Wu wPRF and the Goldreich-Applebaum-Raykov wPRF. Putting everything together\, we obtain public-key PCFs with a throughput of 15k-40k OT/s\, which is of a similar order of magnitude to the state-of-the-art interactive PCFs and about 4 orders of magnitude faster than state-of-the-art public-key PCFs.
URL:https://www.loria.fr/event/ssl-loria-security-seminar-geoffroy-couteau/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20241105T130000
DTEND;TZID=Europe/Paris:20241105T140000
DTSTAMP:20241104T103114Z
CREATED:20241016T073447Z
LAST-MODIFIED:20241104T103114Z
UID:26770-1730811600-1730815200@www.loria.fr
SUMMARY:Formal Methods Seminar
DESCRIPTION:The Formal Methods Seminar takes place every first Tuesday of the month at 13:00 in room A008. \nThe next seminar will take place on Thursday\, November 5th. \nSpeaker: Julie Cailler (Veridis) \nTitle: SC-TPTP : étendre le format TPTP pour les preuves basées sur les séquents \nVisit the Formal Methods Seminar website.
URL:https://www.loria.fr/event/formal-methods-seminar/
LOCATION:A008
CATEGORIES:Séminaire
END:VEVENT
END:VCALENDAR