Thomas Guyet - Inria
Researcher in Computer Science - Inria/AIstroSight

News

  • 12-13/12/2024 -- "Folles journées des anévrismes intracraniens": Modelling care pathways from observational data
  • 03/10/2024 -- Inria x France Biotech: I'll present some synthetic data generation we are developping in AIstroSight Team including a synthetic SNDS data generator.
  • 24/09/2024 -- We are pleased to start the QuickRare project, funded by Shapemed, in which we will develop tools to accelerate the access to diagnosis for rare diseases.
  • 10/09/2024 -- I participated to the seminar "Shaping the future of AI and Health" organized in Oxford by the Maison Française. This event has been supported by the ambassy of UK, Canada and France.
  • 01/06/2024 -- Add EGC2025 in your agenda! I'm very proud to chair the french conference on knowledge extraction and data management this year. See you in Strasbourg!!
  • 14/06/2024 -- We now provide a light and simple implementation of VCNet: the efficient generator of realistic counterfactuals! Check it out!
  • 12/06/2024 -- Arnaud D. joins AIstroSight to contribute to a library of care pathway analysis!
  • 03/06/2024 -- I participated to EINS: a multidisciplinary summer school on informatic health. Have a look at my presentation on modelling care pathways from administrative health data.
  • 28/05/2024 -- GdR MADICS Symposium: I presented the french insurrance database (presentation) to the TIDS workshop.
  • 30/04/2024 -- Our article SWoTTeD: an extension of tensor decomposition to temporal phenotyping has been published in the journal of Machine Learning. The code is available here: SWoTTeD.
  • 27/02/2024 -- The git repository of VCNet is now open. You can test the generation of realistic counterfactuals explanation conceived by V. Guyomard.
  • 19/02/2024 -- Ismail started his PhD in collaboration with F. Fessant and T. Bouadi on robust explanations.
  • 23/11/2023 -- Victor will defend his PhD entitled ``Generation of robust and realistic counterfactual explanations for individual decisions''. The external jury members are F. Giannotti, M.-J. Lesot and S. Gambs.
  • 13/11/2023 -- I'm honored to make an invited presentation about declarative sequential pattern mining in Answer Set Programming during the 32th ILP/IJCLR conference in Bari/Italy.
  • 01/11/2023 -- Andréa and Thibaut just started their PhD under my co-supervision. I'm really thrilled to start new exciting projects for talented candidates.
  • 31/07/2023 -- Our join book "Chronicles: Formalization of a Temporal Model" with P. Besnard is now available. It presents the formal details of the temporal model of Chronicles. For practical use, reach out the PyChronicle python packages.
  • 15/06/2023 -- Victor Guyomard will present his work on robust counterfactual generation and a demonstration of the interface to analyse counterfactuals (for tabular data) during ECML 2023
  • 07/06/2023 -- Kick-off of the PEPR Digital Health is also the official start of the SAFEPAW project.
  • 21/04/2023 -- The 8th edition of the AALTD workshop will be held by ECML. Consider to submit to the workshop or to participate to the competition.
  • 13/04/2023 -- I gave an invited talk during the workshop on EHealth \& Ethics organized by the Da Vinci Research Center (slides are here.
  • 06/01/2023 -- Presentation of the DRIAS project to the "rendez vous de la MITI"
Spring

Bio

I am an full researcher at the Inria Center of Lyon, since 2021. I am part of the AIstroSight team which investigates digital science tools to improve drug design through their application ranging from biological processes understanding to epidemiology. I am currently co-holder of an Inria/APHP chair AI-Raclès on high-dimensional data for care pathways analysis. I am also Vice-President of the French Assocation for Artificial Intelligence (AFIA).
From 2007 to 2021, I was assistant professor at Institut Agro -- AGROCAMPUS OUEST and was working in the IRISA laboratory (LACODAM Team) in Rennes, chair of the Health and Biology cross-cutting research theme. In 2020, I defended my habilitation "Enhancing sequential pattern mining with time and reasoning". I received my PhD thesis in Computer Science in 2007 from National Polytechnic Institute of Grenoble. During my Phd ("Collaborative interpretation of time series. Application to intensive care data"), I worked in the Grenoble Informatic Laboratory (LIG) in a wide range of the artificial intelligence domain (cognitive science, data mining, time series analysis and health data) and in the TIMC Laboratory in a team that works in the applied field of cardio-respiratory physiology (PRETA).

Research interests

My research area is artificial intelligence (AI) with a multidisciplinary approach. I am particularly interested in discovering temporal patterns in semantically rich datasets. This involves the development of innovative methodologies, including PDE discovery, time series analysis, timed-tensor decomposition, timed sequence clustering. My areas of application are related to health (care pathways analysis, cell dynamics understanding). Formally, I was also interested in agronomic science (landscapes analysis, remote sensing)

My first research direction is the temporal and spatial pattern discovery. Data from the observation of living systems (agricultural and medical systems) have a strong spatial or temporal dimension. But the spatial and temporal information is often underutilized in the data mining or machine learning algorithms. The challenge lies in identifying new kind of temporal/spatial patterns that have valuable properties to make possible their extraction by complete and correct algorithms.
My second research direction aims at better including the user(s) in the loop of knowledge discovery. Specifically, I am interested in implementing interactive systems to support users in their exploration process of large datasets. This leads me to contribute to multidisciplinary project with colleagues in philosophy (epistemology, ethic), law studies or cognitive sciences.

Current supervisions

  • Eric Pardoux, PhD candidate (2021-2024), with M Laerke (CNRS, co-dir)
  • Hana Sebia, PhD candidate (2022-2025), with H. Berry (co-Dir)
  • Andréa Ducos, PhD candidate (2023-2026), with A. Denizot (co-Dir)
  • Thibaut Peyric, PhD candidate (2023-2026), with A. Crombach from Beagle Team (co-Dir)
  • Ismail Bachchar, PhD candidate (2024-2027), with F. Fessant (Orange Lab, Co-dir) and T. Bouadi.

Former PhD supervisions

  • Victor Guyomard, PhD candidate (2020-2023), with F. Fessant (Orange, co-Dir), A. Termier (co-dir) and T. Bouadi
  • Nicolas Sourbier, PhD candidate (2019-2022), with M. Pelcat (Dir), F. Majorczyk and O. Gesny
  • Johanne Bakalara, PhD candidate (2018-2021), with O. Dameron, E. Oger (co-Dir) and A. Happe
  • Colin Leverger, PhD candidate (2017-2020), with A. Termier (Dir), S. Malinowski, R. Marguerie (Orange)
  • Clement Gautrais, PhD candidate (oct 2015-2018), with A. Termier (Dir), R. Quiniou, P. Cellier
  • Yann Dauxais, PhD candidate (feb 2015-2018), with D. Gross-Amblard and A. Happe (EA REPERES)

Publications

Projects

Current projects in which I am strongly involved in

  • QuickRare (2024-2026, ShapeMed) in collaboration with Naoual Bakrin (HCL/R2D). This project addresses the problem of reducing the delay to access a diagnosis of rare disease.
  • AI-RACLES (2019-, Chaire APHP/Inria/Central Supelec) in collaboration with Pr. E. Audureau (APHP), A. Tennenhaus (CS) and L. Le Brusquet (CS). This project aims at using high-dimensional data analysis techniques to explore APHP datamart.
    Within the project OPTISOINS (CSE-200001) (Y.-L. Nguyen, P. Pinson and N. Beeker), we investigate the care pathway of patient being hospitalized for lung cancers.
  • ED-AIM (2021-2024, CNRS Prime 80) in collaboration with P. Marty (CNRS/MFO), M. Laerke (CNRS/ENS Lyon). This project aims at investigating an ethical framework for artificial intelligence in medecine.
  • SafePaw (2023-2027, PEPR Digital Health). Multidisciplinary project on multi-view care pathway analysis.

Past projects

  • DRIAS (2020, CNRS MITI) This project explore the notion of public laws in the context of artificial intelligence technologies applied to health (and more especially the accountability of public institutes) in collaboration with IDPSP, REPERES, CERNA, INSPE.
  • PEPS (Pharmaco-Epidemiology of Health Product) (2015-2019, ANSM) The main aims of this project is to make pharmacoepidemiological studies from the medico-administrative databases. These databases contains administrative information about the reimbursement of the 65 millions of french insured. I'm responsible of the scientific program of the IRISA partner. Our aim is to develop temporal pattern mining techniques for care pathway analysis.
  • Hyptser (Hybrid models for predicting time series) (2017-2019, PGMO/IRSDI) collaborative work with S. Malinowski (LinkMedia) and V. Lemaire (Orange Labs).
  • Negative sequential pattern analysis (2017, EDF) collaborative work with the french electricity provider on negative sequential pattern analysis.
  • SePaDec (Sequential Patterns with Declarative Approaches) (2015-2016, Région Bretagne) The aim of this project is to explore declarative approaches, especially Anwser Set Programming to extract sequential patterns. This project will be applied to deeply explore small sets of care pathways. The project is being carried out with B. Negrevergne.
  • PayTal (2012/2015, PPD2) Urban sprawling and landscape relations analysis.
  • PayOTe (2009-2012, turns into an INRA Network) Characterization and simulation of realistic agricultural landscapes.
  • Couplage observation/simulation pour l'aide à la décision dans des syst&eagrave;mes complexes environnementaux (2014, RTR RISC-E) This project aimed at analysing the opportunity offered by artitificial intelligence and pattern mining to analyse remote sensing images.
  • SéSur - Security and monitoring in data streams (2007-2008, Inria) We developed on-line adaptative diagnosis system to analyse a stream of web-logs.

Dissimination

To contact me

Centre de recherche Inria de Lyon
Campus La Doua Villeurbanne Bâtiment CEI-2
56, Boulevard Niels Bohr - CS 52132
69603 Villeurbanne – France

Mail : thomas_dot_guyet_at_inria.fr
Phone : +33(0) 4 72 43 74 96
Devs: Inria GitLab, GitHub
Social networks: ORCID logo, LinkedIn logo