Description du sujet de thèse

The dynamics of marine ecosystems is the result of numerous interactions between species, directly or indirectly determining several services: regulation of climate and water quality, supply of marine products… Understanding these interactions is therefore a key to determining the temporal evolution of such ecosystems. Phytoplankton, a compartment at the basis of food webs, is extremely sensitive to environmental variations and can therefore be used as an indicator. The creation of models to represent the mutual influences between different species (such as competition) is therefore of great interest. Previous works [Karasiewicz et al., 2018] were able to establish the abiotic ecological niche of microalgae species based on multivariate statistical methods, and suggested the prominent role of species interactions but were unable to characterize it.

Based on existing sampling data [SRN, 2017], this thesis aims to propose a method for automatic construction of species interaction models. To this end, the preferred approach will be Learning From Interpretation Transition [Ribeiro et al., 2021], a learning approach that produces logical programs from observations, while ensuring their explainability, unlike other widespread statistical approaches. From these logic programs, an interaction graph will be extracted to allow either a temporal simulation of the model or the study of annual fluctuations in populations of phytoplankton species using abstract interpretation methods. Theoretical improvements of the learning method are also expected, such as the evaluation of an approximate algorithms or allowing to learn from the Most Permissive Semantics [Paulevé et al., 2020].

[Karasiewicz et al., 2018] Stéphane Karasiewicz, Elsa Breton, Alain Lefebvre, Tania Hernández Fariñas, Sébastien Lefebvre. Realized Niche Analysis of Phytoplankton Communities Involving HAB: Phaeocystis Spp. as a Case Study. Harmful Algae 72, 2018. https://doi.org/10.1016/j.hal.2017.12.005
[SRN, 2017] SRN - Regional Observation and Monitoring program for Phytoplankton and Hydrology in the eastern English Channel: 1992-2016 dataset (2017). SEANOE. https://doi.org/10.17882/50832
[Ribeiro et al., 2021] Tony Ribeiro, Maxime Folschette, Morgan Magnin, Katsumi Inoue. Learning any memory-less discrete semantics for dynamical systems represented by logic programs. Machine Learning 11-12, 2021. https://doi.org/10.1007/s10994-021-06105-4
[Paulevé et al., 2020] Loïc Paulevé, Juraj Kolčák, Thomas Chatain, Stefan Haar. Reconciling qualitative, abstract, and scalable modeling of biological networks. Nature Communications 11(4256), 2020. https://doi.org/10.1038/s41467-020-18112-5

Contexte de travail

The person who is recruited will join the BioComputing group of the CRIStAL laboratory, and will be located in the ESPRIT building of the Cité Scientifique campus in Villeneuve-d'Ascq. However, this thesis is part of an interdisciplinary collaboration with the LOG laboratory in Wimereux. The person will thus be supervised by specialists both in computer science (qualitative modeling, automatic learning by inductive logic programming) and ecology (planktonic ecosystems). A continuous discussion between the two disciplines will be set up, in the form of weekly meetings and on-site visits, allowing to regularly validate the progress of the work, the relevance of the methods, and the results.

Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.