Apprentissage

...

Head of Team : Paul Honeine

The main objective comprises modeling and statistical learning studies designed to grasp data diversity (dimensionality, structures, non-stationarity), as well as the nature of the expected solutions (prior knowledge). 

Strategy

  • Multi-source, multi-view, multi-modal learning
  • Semi-supervised or weakly supervised learning
  • Transfer learning, domain adaptation, optimal transport
  • Robust machine learning and trusted AI
  • Deep neural networks
  • Recurrent neural networks
  • Generative adversarial networks
  • Graph analysis, graph matching, metrics on graphs
  • Graph-based models, multi-scale representations
  • Sparse models, dictionary learning and tensor decomposition
  • Information fusion, ensemble learning, random forests
  • Spatial and temporal information modeling
  • Federated learning and serious games
  • Apps

    • Handwriting recognition
    • Analysis of handwritten documents
    • Analysis of old documents
    • Information retrieval
    • Medical imaging
    • Analysis of sports gestures
    • Analysis and synthesis of molecules
    • Analysis of road scenes
    • Road traffic analysis
    • Risk management
    • Historical spatial dynamics
    • Analysis and description of audio scenes
    • Time series analysis

    Partner

    International Collaborations

      National Academic Collaborations
      Academic Collaborations in Normandy
      Industrial Partners

      Projects

      ODD : Online Deep anomaly Detection (ANR, 2023-…)

      Chaire IA RAIMO : A Road toward safe Artificial Intelligence in MObility (ANR, 2020-2024)

      DynaTeam Dynamique de coordination d'équipe (ANR, 2023-…)
      Partners : CETAPS
      Équipe RI2C du LITIS

      FAMOUS : Fair Multimodal Learning (ANR, 2023-…)
      Partners : 
      LISLHCINTEuranova

      FINLAM : Foundation INtegrated models for Libraries Archives and Museum (ANR, 2023-…)
      Partners : TEKLIABnF

      SHARP Machine Learning for Safe Vehicle Charging Points (ANR, 2023-…)
      Partners :
       Citeos Solutions Digitales (VINCI Energies)GREYC

      CoDeGNN : Convolution and Decimation for Graph Neural Networks (ANR, 2021-2025)
      Partners : GREYC
      LIFAT

      MediSeg : Deep medical image segmentation: what’s next? (ANR, 2021-2025)
      Partners : LMIImVIA
      Équipe QuantIF du LITIS

      MultiTrans : Gradual Multi Transfer Learning for Safe Autonomous Driving (ANR, 2021-2025) Partners : Équipe STI du LITIS, I3SValeo.ai 

      EXO-POPP : Optical Extraction of Handwritten Named Entities for the Marriage Certificates of the Population of Paris (1921-1946) (ANR, 2021-2024)
      Main partner : LARHRA

      CATCH : Automatic Understanding of Human Sensors Testimonials (ANR, 2021-2023)
      Partners : SaagieAtmo Normandie

      LabCom L-Lisa : Common Lab LITIS-Saagie (ANR, 2021-2025)
      Partner : Saagie

      SCHISM : Supporting CHemoinformatics via Interactive unsupervised and Semi-supervised data Mining (FEDER, 2020-2024)

      Partners : GREYCCERMN

      DeepART : Medical Image collection, segmentation and anonymization for DEEP learning in Adaptive Radiation Therapy (2020-2022)
      Partners : Centre François BaclesseCentre Henri BecquerelGREYC

      WeSmile : Weakly supervised medical image segmentation (PHC Van Gogh, 2020-2022)

      Partner : Eindhoven TU

      AMCAS : Machine Learning for Understanding Secondary Atomization (2020-2022)
      Partner : CORIA

      THESIS : THermoplastic Erosion Shield for new generation Ice protection System (2019-2022)
      Partners : Zodiac Aerosafety Systems (nowadays SAFRAN), Analyse and Surface, Dedienne Multi-plasturgy Group, IRSEEMGPMCORIA

      ESCAPE SG : Serious game for evacuation management (2019-2023)

      Partners : UMR IDEESUMI UMMISCO

      APi : Taming the Beast of the Preimage in Machine Learning for Structured Data : Signal, Image and Graph (ANR, 2018-2023)

      Partners : LTCI, LIG

      PAUSE APi : in complement to ANR APi, this funding is from the national program for the urgent aid and reception of scientists in exile (PAUSE, from the Collège de France)

      NormanD'eep : Deep learning and application (2018-2022) 
      Partner : GREYC

      ASTURIAS : Newspaper image analysis (2018-2022) 
      Partners : GREYCMIND group of LITIS

      DeepInFrance : Machine learning with deep neural networks (ANR, 2018-2022)
      Partners : INRIA GrenobleI3SGREYCLIF-AMULIP6

      LEAUDS : Learning to understand audio scenes (ANR, 2018-2022)
      Partners : NetatmoINRIA Nancy - Grand Est

      ESCAPE : Exploring by Simulation Cities Awareness on Population Evacuation (ANR, 2017-2022)
      Partners : UMR IDEESUMI UMMISCO

      DynACEV : Dynamics of Learning: Behaviour and Lived Experiences. The role of exploratory strategies (ANR, 2017-2022)
      Partners : CETAPSISSUL (Lausanne), MIPSOPESES (Nouvelle-Zélande), SPARC (UK) 

      OATMIL : Optimal Transport and Machine Learning (ANR, 2017-2022) 
      Partners : IRISALagrangeTechnicolor 

      HBDEX :  Analysis of historical financial data (ANR, 2017-2022)
      Partners : PSECAMSIRISA

      POPP : Project for the oceration of the Paris population census (1921-1946) (2020-2021) - voir démonstrateur ici
      Partner : LARHRA

      EurHisFirm :  The European project aims at designing a world-class research infrastructure to collect, merge, extract, collate, align and share detailed historical high-quality firm level data for Europe (H2020 EU, 2018-2021)
      Partners : Consortium of 16 European partners 

      Read E-Doc : Development of a system for extracting specific alphanumeric information from dematerialized container shipment slips (2019-2021)
      Partner : Toshiba


      Our PhD students in industry (CIFRE programs) with companies DataHertzLokadTEKLIAWeinig LuxscanActemiumSOLYSTICItekube

      Software


      ...

      PIVAJ

      PIVAJ (Plateforme d'Indexation et de Visualisation d'Archives de Journaux) is a software suite for automatic analysis and display of digitized historical newspapers.The offline part extracts sections, articles and text from the digitized images, to build structured METS/ALTO files.

      En savoir plus
      ...

      DocExplore

      DocExplore is a free software suite to build and display augmented books from digitization. DocExplore is WySiWyG, no XML knowledge required, and works on Windows, MacOs and Linux.

      En savoir plus
      ...

      Products and tools from the Machine Learning group

      En savoir plus