Personal home page of Grégoire Sergeant-Perthuis
Email: gregoireserper@gmail.com , CV , ResearchGate
Interests: Categorical probability theory, Statistical Physics, Artificial Intelligence, Biology, Neurosciences
My PhD thesis: Full text ,
Postdoctoral researcher (Mathematics, University of Artois) and CSO @Projective Consciousness Robotics (Events)
Categorical formulation of Gibbs states
In the classical formulation of Gibbs states ('Gibbs Measures and Phase Transitions' H.-O. Georgii) one must specify a constrained collection of probability kernels called a specification. One can show that specifications are a particular case of couples of functors/cofunctors from a poset to the category of measurables spaces/Markov kernels and that Gibbs states are sections of these couples (Chapter 8 PhD thesis ). A particular case of acyclic couples of functors/cofunctors (for the derived limit functor) are exhibited, that we call decomposable; they have strong motivations in pobability and statistical physics. A resolution in these acyclic objects helps compute the cohomology groups. Presentation slides (PhD Defence)
Research work: (Pre-)Publications
[1] "Bayesian/graphoid intersection property for factorisation models", arXiv:1903.06026 , 2021.
[2] "Intersection property and interaction decomposition", arXiv:1904.09017 , 2021.
[4] "Around intersection property and interaction decomposition", 2021.
[5] "Extra-fine sheaves and interaction decompositions" (with D. Bennequin, J. P. Vigneaux, O. Peltre), arXiv:2009.12646 , 2020.
Key words: categorical probability, Gibbs states
Adaptive systems in biology
One of the frameworks for understanding how `brains' function is the Bayesian Brain Hypothesis. It states that, in order to preserve their integrity, adaptive systems have evolved so that they can predict the behaviour of their environment. How do they do so? By making hypothesis on the state of the world and uptdating them through observation. In this work we explore how such behaviours appears in activity of biological system without central nervous systems for example at cellular and intracellular scale.
Research work: (Pre-)Publications
Research work: (Pre-)Publications
[1] "Modeling the subjective perspective of consciousness and its role in the control of behaviours" (with D. Rudrauf, O. Belli, Y. Tisserand, G. Di Marzo Serugendo), arXiv:2012.12963 , Journal of Theoretical Biology, 2022.
[2] "Combining the Projective Consciousness Model and Virtual Humans to assess ToM capacity in Virtual Reality: a proof-of-concept" (with D. Rudrauf, Y. Tisserand, T. Monnor, O. Belli), arXiv:2104.07053 , 2021.
Talks
[1] 'Regionalized optimisation and message passing algorithms', Séminaire du Centre de Recherche en Informatique de Lens (CRIL), Lens, 31 March 2022
[2] 'Regionalized optimisation and message passing algorithms', Workshop 'Geometrie de Wassertein', Laboratoire de Mathematique de Lens, at Lens, 17 December 2021
[3] 'Active inference with perspective taking', Workshop 'Geometrie de Wassertein', Laboratoire de Mathematique de Lens, at Lille, 17 December 2021
[4] 'The role of consciousness in social active inference and the emergence of adaptive and maladaptive behaviors', XXIX CONGRESSO NAZIONALE, SIPF, at Palermo, 1 October 2021.
[5] 'The role of consciousness in social active inference and the emergence of adaptive and maladaptive behaviors', Seminario del Istituto di Fisiologia Umana, Istituto di Fisiologia Umana, University of Palermo, 29 September 2021.
[6] 'Modelling consciousness using projective geometry', Seminar on consciousness, LPNC, at Grenoble, 21 September 2021.
The Conferences and Workshops I helped organize:
[1]'Geometrie de Wassertein', Laboratoire de Mathematique de Lens, at Lens 16-17 December 2021
Supervision
Rida Lali (Master 1 ENS-Paris-Saclay) : 'Active inference for emotional agents within the PCM' (at Lab for MulTimodal Modelling of Emotion and Feeling, University of Geneva)
Robin Sobczyk (Licence 3 ENS-Paris-Saclay): 'Use of heuristics for improving computation times of the Projective Consciousness Model' (at Lab for MulTimodal Modelling of Emotion and Feeling, University of Geneva)
Teaching
Statistique, Master 1, University of Paris
Analyse de données, Master 1, University of Paris
Statistiques et simulations probabilistes Licence 3, Université de Paris
Statistiques inférentielles, Licence 1, IUT Paris Descartes
Probabilité et simulations 1, Licence 1, IUT Paris Descartes
Analyse de Hilbert et de Fourier, Licence 3, Université Paris Diderot