Home » HDRs

HDRs 2025



  Evgeny Gurevsky, Quelques contributions à la résolution de problèmes d'optimisation en présence d'incertitude, de modularité ou de contrôlabilité
Quelques contributions à la résolution de problèmes d'optimisation en présence d'incertitude, de modularité ou de contrôlabilité
Auteur : Evgeny Gurevsky
Manuscript :


Mots-clés : Operations researchProduction systemsModularityReconfigurabilityBalancingUncertaintyRobustnessStability radiusRiskControllabilityMILPPre-processingConstraint generationComplexity
Résumé



Date de soutenance : 25-03-2025
Président du jury : Nathalie Bostel
Jury :
  • Christian Artigues [Rapporteur]
  • Jean-Charles Billaut [Rapporteur]
  • Safia Kedad-Sidhoum [Rapporteure]
  • Olga Battaïa
  • Nadjib Brahimi
  • André Rossi
  • Alexandre Dolgui

  Kandaraj Piamrat, From network management towards network analytics: a decade journey of research study
From network management towards network analytics: a decade journey of research study
Auteur : Kandaraj Piamrat
Manuscript :


Mots-clés : Network ManagementNetwork AnalyticsQuality of ExperienceMachine Learning
Résumé

This manuscript explores the evolution of network management and analytics, highlighting key concepts, technologies, and challenges in these domains. In network management, there has been a major shift towards prioritizing user-centric Quality of Experience (QoE), driving significant adaptations. We have investigated both centralized and decentralized approaches for QoE-aware routing and resource allocation across wireless mesh networks, home networks, and sensor networks. Additionally, we have explored cross-layer designs that integrate routing (network layer), application-layer prioritization, and MAC-layer queuing, particularly for optimizing video streaming. In network analytics, we have applied a range of machine learning techniques, including unsupervised learning for network slicing and vehicular user behavior analysis, supervised learning for traffic classification using ensemble models and federated learning, and semi-supervised learning to address label scarcity through stacked autoencoders and federated learning. Furthermore, we have introduced a framework for learning across the Cloud-Edge-IoT continuum, integrating hierarchical federated learning with spiking neural networks to tackle challenges related to scalability, privacy, and energy efficiency. Collectively, these efforts advance the state of the art in network management and analytics, providing comprehensive methodologies to enhance network performance, security, user experience, and offer deeper insights into present and future communication networks.


Date de soutenance : 12-02-2025
Président du jury : Stefano Secci
Jury :
  • Carla Fabiana Chiasserini
  • Yacine Ghamri-Doudane [Rapporteur]
  • Claud Jard
  • Yusheng JI
  • Adlen Ksentini
  • Rami Langar [Rapporteur]

Copyright : LS2N 2017 - Mentions Légales - 
 -