DENIS Christophe
Research activity
The notions of information and conviviality in deep neural networks â or what about "Explainable AI" or "Trust AI"?The thunderous return of neural networks occurred in the sublime Florentine setting in 2012 during a renowned international computer vision conference. As for several years, the participants of this conference were invited to test their image recognition techniques. Geoffrey Hinton's team from the University of Toronto was the only one using deep neural networks: it outperformed the other competitors in two out of three categories of the competition. The audience was stunned by the impact of the reduction in prediction error, a factor of three, while the algorithms based on the expertise of the researchers differ by a few percent. Other computational scientific disciplines, like computational fluid dynamics, geophysics, and climatology, have also started to use deep learning methods to predict phenomena which are difficult to solve with a classical hypothetical deductive approach.
We argue that systematically explaining deep learning to all its users is not always justified, could be counterproductive and even raises ethical issues. For example, how to assess the correctness of an explanation that could even be unintentionally permissive or even manipulative in a fraudulent context? There is therefore a need to revisit the theory of information (Fisher, Shannon) and the philosophy of information (eg. Floridi) in the light of deep learning. This information will allow certain users to produce their own reasoning (surely an abductive one) rather than receiving an explanation.
Last but not least, should we trust a machine learning model? Trust means handing over something valuable to someone, relying on them. The corollary is that "the person who trusts is immediately in a state of vulnerability and dependence", and all the more and all the more so on the basis of an explanation whose correctness is difficult to assess. Last but not least, we strongly believe that using human relationship terms, like trust or fairness in the context of machine learning, necessarily induces anthropomorphism, whose bad effects could be addiction (Eliza effect) and persuasion rather than information. In contrast, our philosophical and mathematical research direction tries to define conviviality criteria in machine learning based on Ivan Illich's thought. According to Illich, a convivial tool must have the following properties: âą it must generate efficiency without degrading personal autonomy; âą it must create neither slave nor master; âą it must widen the personal radius of action. As presented in the last part of the talk, neural differential equations, by providing trajectories rather than predictions, seem to be an efficient mathematical formalism to implement convivial deep learning tools.
One past PhD student (2024) at Sorbonne University
- 2024
- BAYET Théophile : Caractérisation de l'inclusivité des systÚmes de vision par ordinateur basés sur l'apprentissage profond pour les pays du Sud .
2005-2023 Publications
-
2023
- Ch. Denis : “Lâapport de la philosophie et de lâĂ©pistĂ©mologie dans la mise en place dâune politique de cybersĂ©curitĂ© en santĂ©”, (2023)
- Ch. Denis : “Cadre mĂ©thodologique pour assurer la spĂ©cification dâun simulateur par apprentissage machine profond en vue de sa validation”, (2023)
- D. ROTIMBO MBOUROU, D.‑F. Bilenga Moukodouma, Ch. Nkoulembene, Ch. Denis : “Apports de la boĂźte Ă outils Advanced Data Analysis (AdvDatAna) pour le traitement automatisĂ© des bases de donnĂ©es mĂ©tĂ©orologiques : cas du parc de la LopĂ©.”, Colloque Scientifique International, Abidjan, CĂŽte d'Ivoire (2023)
- D.‑F. Bilenga Moukodouma, Ch. Denis, D. ROTIMBO MBOUROU, Ch. Nkoulembene : “Proposition dâune nouvelle approche dâestimation dâĂ©lĂ©phants basĂ©e sur les algorithmes de lâIntelligence Artificielle dans les forĂȘts gabonaises”, JournĂ©e de la Recherche Technologique, de l'Innovation et du DĂ©veloppement durable, Libreville, Gabon (2023)
- Th. Bayet, Ch. Denis, A. Bah, J.‑D. Zucker : “Les dĂ©fis du glissement de contexte gĂ©ographique”, JournĂ©e RĂ©silience et IA - Plate-Forme d'Intelligence Artificielle, Strasbourg, France (2023)
- Ch. Denis, J. Nicogossian : “Du gĂšne Ă lâoctet : la communication phygitale pour une utilisation responsable de lâIntelligence Artificielle dans le domaine mĂ©dical”, Dalloz IP/IT : droit de la propriĂ©tĂ© intellectuelle et du numĂ©rique n°3, pp. 145-150, (Ăditions Dalloz [2016-....]) (2023)
-
2022
- N. Demeure, C. Chevalier, Ch. Denis, P. Dossantos‑Uzarralde : “Algorithm 1029: Encapsulated Error, a Direct Approach to Evaluate Floating-Point Accuracy”, ACM Transactions on Mathematical Software, vol. 48 (4), pp. 47, (Association for Computing Machinery) (2022)
- Ch. Denis : “Peut-on avoir confiance Ă lâIntelligence Artificielle pour nous aider Ă contrer les effets du changement climatique ?”, Fete de la Science - UniversitĂ© de Rouen, Rouen, France (2022)
- Th. Bayet, Ch. Denis, A. Bah, J.‑D. Zucker : “Distribution Shift nested in Web Scraping : Adapting MS COCO for Inclusive Data”, ICML Workshop on Principles of Distribution Shift 2022, Baltimore, United States (2022)
- Ch. Denis, F. Varenne : “InterprĂ©tabilitĂ© et explicabilitĂ© de phĂ©nomĂšnes prĂ©dits par de lâapprentissage machine”, Revue Ouverte d'Intelligence Artificielle, vol. 3 (3-4), pp. 287-310, (Association pour la diffusion de la recherche francophone en intelligence artificielle) (2022)
- Ch. Denis : “Le pĂ©riple de lâĂ©thique de lâintelligence artificielle dans les rĂ©volutions en cours des systĂšmes de soins”, chapitre de Intelligence Artificielle. Vivre Avec.Vers Une Nouvelle Architecture Du Monde, (ISBN: 978-2-8224-0801-1) (2022)
- Ch. Denis : “Esquisses philosophiques autour de la comprĂ©hension de phĂ©nomĂšnes complexes avec des outils de prĂ©diction basĂ©s sur de lâapprentissage machine”, EGC - ConfĂ©rence francophone sur l'Extraction et la Gestion des Connaissances - Atelier Explain'AI, Blois, France (2022)
- J. CĂĄrdenas ChapellĂn, Ch. Denis, H. Mousannif, Ch. Camerlynck, N. Florsch : “Magnetic Anomalies Characterization: Deep Learning and Explainability”, Computers and Geosciences, (Elsevier) (2022)
-
2021
- Ch. Denis : “Enjeux scientifiques et Ă©pistĂ©mologiques de lâexplicabilitĂ© de la prĂ©diction de phĂ©nomĂšnesphysiques par de lâapprentissage machine profond”, VIIIe congrĂšs de la SociĂ©tĂ© de Philosophie des Sciences, Mons, Belgium (2021)
- Ch. Denis : “Le pĂ©riple de lâĂ©thique de lâIntelligence Artificielle dans la rĂ©volution en cours des systĂšmes de soins”, Droit, SantĂ© et SociĂ©tĂ© [Journal de mĂ©decine lĂ©gale, droit mĂ©dical, victimologie, dommage corporel. SĂ©rie E], vol. 3 (3), SantĂ© et intelligence artificielle Quelle(s) rĂ©volution(s)âŻ?, pp. 17-21, (Eska [2014-....]) (2021)
- Th. Bayet, T. Brochier, Ch. Cambier, A. Bah, Ch. Denis, N. Thiam, J.‑D. Zucker : “A Machine Learning approach to improve the monitoring of Sustainable Development Goals : a case study in Senegalese artisanal fisheries”, CNIA 2021 : ConfĂ©rence Nationale en Intelligence Artificielle, Bordeaux (virtuel), France, pp. 30-37 (2021)
- J. CĂĄrdenas ChapellĂn, Ch. Denis, H. Mousannif, Ch. Camerlynck, N. Florsch : “RĂ©seaux de Neurones Convolutifs pour la CaractĂ©risation dâAnomalies MagnĂ©tiques”, Actes CNIA 2021, Bordeaux (en ligne), France, pp. 84-90 (2021)
- J. Cardenas Chapellin, Ch. Denis, H. Mousannif, Ch. Camerkynck, N. Florsch : “RĂ©seaux de neurones convolutifs pour la caractĂ©risation dâanomalies magnĂ©tiques.”, Colloque GEOFCAN, A distance, France (2021)
-
2019
- Th. Anglade, Ch. Denis, Th. Berthier : “A novel embedding-based framework improving the User and Entity Behav- ior Analysis”, (2019)
- Ch. Denis, J. Nicogossian : “Du gĂšne Ă lâoctet : la communication phygitale pour une utilisation responsable de lâIntelligence Artificielle dans le domaine mĂ©dical”, (2019)
- Ch. Denis, F. Varenne : “InterprĂ©tabilitĂ© et explicabilitĂ© pour lâapprentissage machine : entre modĂšles descriptifs, modĂšles prĂ©dictifs et modĂšles causaux. Une nĂ©cessaire clarification Ă©pistĂ©mologique”, Actes ConfĂ©rence Nationale en Intelligence Artificielle (CNIA), PFIA 2019 (https://www.irit.fr/pfia2019/wp-content/uploads/2019/07/actes_CNIA_PFIA2019.pdf), Toulouse, France, pp. 60-68 (2019)
- Ch. Denis : “Towards an explainable and convivial AI based tools: Illustration on medicine applications”, http://www.chistera.eu/christophe-denis, Tallinn, Estonia (2019)
- Ch. Denis, J. Nicogossian : “ExplicabilitĂ© et convivialitĂ© dâoutils de diagnostic mĂ©dical basĂ© sur de lâIA”, JournĂ©es plĂ©niĂšres du GdR-IA, Groupe de travail ExplicabilitĂ©, OrlĂ©ans, France (2019)
- Ch. Denis : “Collaboration intelligence humaine et intelligence artificielle en vue dâune sociĂ©tĂ© conviviale”, Forum Industriel de l'Intelligence Artificielle (FIIA'2019), Champ-sur-Marne, France (2019)
- Ch. Denis, A. Delaborde, N. Meric : “Table Ronde TRB3 "IA, confiance, sĂ©curitĂ©"”, IANP 2019 â Intelligence Artificielle : Nouvelles Puissances, Paris, France (2019)
-
2013
- S. Montan, J.‑M. Chesneaux, Ch. Denis, J.‑L. Lamotte : “Ătude de la propagation des erreurs dâarrondi dans un code dâhydrodynamique parallĂšle”, ConfĂ©rence d'informatique en ParallĂ©lisme, Architecture et Systeme, COMPAS 2013, Grenoble, France (2013)
-
2012
- S. Montan, J.‑M. Chesneaux, Ch. Denis, J.‑L. Lamotte : “Towards an efficient implementation of CADNA in the BLAS : Example of DgemmCADNA routine.”, 15th GAMM - IMACS International Symposium on Scientific Computing, Computer Arithmetic, and Validated Numerics (SCAN), Novosibirsk, Russian Federation (2012)
- Ch. Denis, S. Montan : “Numerical Verification of Industrial Numerical Codes”, CongrĂšs National de MathĂ©matiques AppliquĂ©es et Industrielles, vol. 35, ESAIM: Proceedings, Guidel, France, pp. 107-113 (2012)
- F. JĂ©zĂ©quel, R. Couturier, Ch. Denis : “Solving large sparse linear systems in a grid environment: the GREMLINS code versus the PETSc library”, Journal of Supercomputing, vol. 59 (3), pp. 1517-1532, (Springer Verlag) (2012)
-
2009
- F. JĂ©zĂ©quel, Ch. Denis, Ph. TrĂ©buchet : “Reliable numerical evaluation of eigenvalues involved in polynomial systems solving”, Computer-assisted proofs - tools, methods and applications, Dagstuhl, Germany (2009)
- Ch. Denis, R. Couturier, F. JĂ©zĂ©quel : “A sparse linear system solver used in a distributed and heterogeneous grid computing environment”, chapter in Parallel Scientific Computing and Optimization, vol. 27, Springer Optimization and Its Applications, pp. 47-56, (Springer) (2009)
-
2008
- Ch. Denis, F. JĂ©zĂ©quel : “Numerical validation of eigenvalues computation in solving polynomial systems”, Proceedings of the 8th Real Numbers and Computers conference, Santiago de Compostela, Spain, pp. 123-131 (2008)
- N. Scott, V. Faro‑Maza, M. Penny Scott, T. Harmer, J.‑M. Chesneaux, Ch. Denis, F. JĂ©zĂ©quel : “E-collisions using e-science”, Physics of Particles and Nuclei Letters [PisĐ'ma v Zhurnal Fizika Elementarnykh Chastits i Atomnogo Yadra / PisÊčma v ĆŸurnal "Fizika Ăšlementarnyh Äastic i atomnogo Ăądra"], vol. 5 (3), pp. 150-156, (MAIK Nauka/Interperiodica ; Pleiades Publishing [2007-....]) (2008)
- R. Couturier, Ch. Denis, F. JĂ©zĂ©quel : “GREMLINS: a large sparse linear solver for grid environment”, Parallel Computing, vol. 34 (6-8), pp. 380-391, (Elsevier) (2008)
-
2007
- N. Scott, M. Penny Scott, Liviu Gr. Ixaru, Ch. Denis : “2DRMP: fast computation of the Slater integrals”, Mathematical and Computational Methods in R-matrix theory, CCP2 workshop proceedings, London, United Kingdom, pp. 70-75 (2007)
- Ch. Denis, F. JĂ©zĂ©quel, N. Scott : “High Performance Computation: numerical music or numerical noise?”, Proceedings of the 8th HERCMA (Hellenic European Research on Computer Mathematics and its Applications) conference, Athens, Greece, pp. 1-8 (2007)
- N. Scott, F. JĂ©zĂ©quel, Ch. Denis, J.‑M. Chesneaux : “Numerical âhealth checkâ for scientific codes: the CADNA approach”, Computer Physics Communications, vol. 176 (8), pp. 507-521, (Elsevier) (2007)
- T. Sadowski, E. Postek, Ch. Denis : “Stress distribution due to discontinuities in polycrystalline ceramics containing metallic inter-granular layers”, Computational Materials Science, vol. 39 (1), pp. 230-236, (Elsevier) (2007)
-
2006
- N. Scott, Liviu Gr. Ixaru, Ch. Denis, F. JĂ©zĂ©quel, J.‑M. Chesneaux, M. Penny Scott : “High performance computation and numerical validation of e-collision software”, International Conference of Computational Methods in Sciences and Engineering 2006 (ICCMSE 2006), vol. 6, Lecture Series on Computer and Computational Sciences, Chania, Crete, Greece, pp. 561-570, (CRC Press) (2006)
- E. Postek, T. Sadowski, Ch. Denis : “Modelling of Metalling of Inter-granular Layers in Polycrystalline Ceramics”, Eighth International Symposium on Brittle Matrix Composites, Warsaw, Poland, pp. 495-505, (Zturek RSI and Woodhead Publ.) (2006)
- J.‑M. Chesneaux, Ch. Denis, F. JĂ©zĂ©quel, N. Scott : “Dynamical strategies for the computation of integrals: an example from the CPC library”, Proc. SCAN2006 conference, Duisburg, Germany, pp. 41-42 (2006)
- S. Contassot‑Vivier, R. Couturier, Ch. Denis, F. JĂ©zĂ©quel : “Efficiently solving large sparse linear systems on a distributed and heterogeneous grid by using the multisplitting-direct method”, Proc. 4th International Workshop on Parallel Matrix Algorithms and Applications, PMAA'06, Rennes, France, pp. 1-2 (2006)
- N. Scott, V. Faro‑Maza, M. Penny Scott, T. Harmer, J.‑M. Chesneaux, F. JĂ©zĂ©quel, Ch. Denis : “e-Collisions Using e-Science”, Mathematical Modeling and Computational Physics, High Tatra Mountains, Slovakia (2006)
- Ch. Denis : “Parallelization of a Numerical Code Predicting the Mechanical Response of a Polycrystalline Ceramics containing Metalling Integral Layers Under Uniaxal Tension”, chapter in Science and Supercomputing in Europe - Report 2005, pp. 547-550, (CINECA), (ISBN: 88-86037-17-1) (2006)
-
2005
- Ch. Denis, E. Postek, T. Sadowski : “A Parallel Numerical Code Predicting the Mechanical Response of a Polycrystalline Ceramic Material Containing Inter-granular Layers”, TAM'O5 (Transnational Access Meeting) Workshop, HPC-Europa, Stuttgart, Germany, pp. 1-2 (2005)
- Ch. Denis, J.‑P. Boufflet, P. Breitkopf : “A Load Balancing Method for a Parallel Application Based on a Domain Decomposition”, IPDPS 2005 - 19th IEEE and ACM Int. Parallel and Distributed Processing Symposium, Denver, Colorado, United States, pp. 1-8, (IEEE Computer Society Press) (2005)
- Ch. Denis, J.‑P. Boufflet, P. Breitkopf, M. Vayssade : “Ăquilibrage en volume de calcul pour une mĂ©thode parallĂšle Ă fronts multiples”, Revue EuropĂ©enne des ĂlĂ©ments Finis, vol. 14 (1), pp. 87-113, (HERMĂS / LAVOISIER) (2005)