Accéder directement au contenu
JD

Jean-Baptiste Durand

Jean-Baptiste Durand - Statistics and stochastic modelling of plant growth
1
Documents
Identifiants chercheurs

Présentation

### Webpage: [http://amap-collaboratif.cirad.fr/pages-chercheurs/?page\_id=32153](http://amap-collaboratif.cirad.fr/pages-chercheurs/?page_id=32153) ### Short biography / Research topics Jean-Baptiste Durand received the engineering degree in applied mathematics and computer science from the "Institut National Polytechnique de Grenoble" [Grenoble INP](http://www.grenoble-inp.fr/grenoble-institute-of-technology-9224.kjsp?RH=INPG) (Grenoble Institute of Technology) in 1999. He received the Ph.D. degree in applied mathematics in 2003 from the University of Grenoble I (France), now [Université Grenoble Alpes](https://www.univ-grenoble-alpes.fr). From 1999 to 2003, he worked on statistical modelling at the French National Institute for Research in Computer Science and Control ([Inria](http://www.inria.fr/en/)), as a member of the [IS2 project](http://www.inrialpes.fr/is2) (Statistical Inference for Health and Industry). Then, he was a Postdoctoral Research Fellow at [AMAP](http://amap.cirad.fr/) "Plant Modelling" mixed research unit, Montpellier (France), with a grant from [CIRAD](http://www.cirad.fr/en). From 2004 to 2022, he was a teaching assistant at [Grenoble INP](http://www.grenoble-inp.fr/92723626/1/fiche___pagelibre/), member of the [Statify](https://team.inria.fr/statify/) team of [Laboratoire Jean Kuntzmann](http://www-ljk.imag.fr/). He was the head of the statistics and data science tracks of the international Master's programme MSIAM <http://msiam.imag.fr/>. He taught the fundamentals of statistics and computational statistics at the "École Nationale Supérieure d'Informatique et de Mathématiques Appliquées de Grenoble" [Ensimag](http://ensimag.grenoble-inp.fr/ecole-nationale-superieure-d-informatique-et-de-mathematiques-appliquees-74488.kjsp?RH=ENSIMAG_FR) (engineering school of applied mathematics and computer science). He was delegated to Inria team [Virtual Plants](http://team.inria.fr/virtualplants/) (Inria Sophia) between September 2009 and August 2011 and to Inria team [Statify](https://team.inria.fr/statify/) between September 2019 and August 2021 . His research interests include computational methods for hidden Markov models, statistical analysis of tree-structured processes, discrete multivariate distributions and applications to signal processing and botany. He has returned to [AMAP](http://amap.cirad.fr/) as a [CIRAD](http://www.cirad.fr/en) researcher since 2022 and now aims at developing statistical models and approaches for the analysis of structured data issued from plant phenotyping. ![](http://amap-collaboratif.cirad.fr/pages-chercheurs/wp-content/uploads/JDURAND_rain_lin.gif) ### Related working groups, research teams and pages - Research teams: - formerly [Virtual Plants](http://team.inria.fr/virtualplants/) (Inria Sophia), now [MOSAIC](https://www.inria.fr/fr/mosaic) in Lyon, see also their [research program](http://www.ens-lyon.fr/RDP/spip.php?rubrique52). - CIRAD [PhenoMEn](https://umr-agap.cirad.fr/nos-recherches/equipes-scientifiques/phenotypage-et-modelisation-des-plantes-dans-leur-environnement-agro-climatique/organisation-de-l-equipe) - mainly, the former M2P2 team (Montpellier) - AGAP [AFEF](https://umr-agap.cirad.fr/nos-recherches/equipes-scientifiques/architecture-et-floraison-des-especes-fruitieres/contexte-et-enjeux) (Montpellier) ![](http://amap-collaboratif.cirad.fr/pages-chercheurs/wp-content/uploads/JDURAND_rain_lin.gif) ### <a name="publis"></a>Publications (See [copyryight notice](http://amap-collaboratif.cirad.fr/pages-chercheurs/?page_id=32360)) #### [Publications on HAL](https://hal.archives-ouvertes.fr/search/index/q/*/authIdHal_s/jean-baptiste-durand) #### PhD thesis (defended the 31st of january 2003) ### Modèles à structure cachée : inférence, estimation, sélection de modèles et applications (in French) ([Compressed PDF](http://amap-collaboratif.cirad.fr/pages-chercheurs/wp-content/uploads/JDURAND_these.pdf.gz)) #### [Habilitation](https://en.wikipedia.org/wiki/Habilitation) thesis (defended the 19th of October 2020) ### Contributions to hidden Markov models and applications to plant structure analysis ([Compressed PDF](http://amap-collaboratif.cirad.fr/pages-chercheurs/wp-content/uploads/JDURAND_habilitation.pdf.gz)) #### Book - [Data Science. Cours et exercices.](https://www.eyrolles.com/Informatique/Livre/data-science-9782212674101) M.-R. Amini, R. Blanch, M. Clausel, J.-B. Durand, E. Gaussier, J. Malick, C. Picard, V. Quéma et G. Quénot. Eyrolles (Éd.), 2018. ![](http://amap-collaboratif.cirad.fr/pages-chercheurs/wp-content/uploads/JDURAND_rain_lin.gif) ### <a name="logiciels"></a>Software [OpenAlea](https://github.com/openalea/) - Author of package [TREE\_STATISTIC](https://github.com/openalea/StructureAnalysis/tree/master/tree_statistic) in [StructureAnalysis](https://github.com/openalea/StructureAnalysis) <!-- ### <a name="phd"></a>PhD Proposal #### Computational methods for hidden semi-Markov models with mixed effects -- application to plant and root branching patterns Profile and required skills required: Skills in statistics (modelling, parametric estimation) and if possible, random processes. Taste for applications and programming. ##### Abstract In the framework of plant development modelling, statistical models can be divided into two categories. The first one, referred to as "genotype x environment", is based on mixed models, which do not account for time dependencies existing in the considered processes. The second category is based on sequence analysis models that are funded on biological models, but currently do not account for genotype or environmental effects. More specifically, we are here focused on hidden semi-Markov models, introduced about 20 years ago (Guédon *et al.*, 2001) to model dynamical aspect of plant structure development. These models allow modellers to account for different development phases of either plants or their components (branches, roots, etc) through hidden states. The PhD proposal aims at including fixed and random effects within this category of models, the former aiming at characterising the effects of targeted covariates (genotype and environment) ant the latter, to account for constraints related to experimental design. The work to be accomplished, beyond model specification, is to develop inference algorithms suited to the specific complexity of these models . This PhD proposal opens avenues in two fields. Firstly, the newly developed family of models will be included as basic components within more global models, which is a major concern in model agronomic methods. Moreover, the methodological advances obtained in the hidden semi-Markov framework will enrich this family of models and offer new possibilities for addressing scientific questions in various domains of application (health, seismology, reliability, ecology, etc). ##### Context and aims The fast development of automatic platforms for plant phenotypic, achieving dynamical measurements of plant characteristics, is now leading to time and spatially dependent data. These data can lead to precious information to guide crop management, provided adequate models are built to analyse them. Most common plant development models are, one the one hand, deterministic approaches based on complex biological models, which hardly account for the observed phenotypic variability. On the other hand, statistical regression models offer possibilities to test genotype and environment effects, but they rely on over-simplistic biological time processes, for example summarized through an unique coefficient. Between both approaches, statistical methods including dynamical aspects of plant development were proposed by Guédon *et al.* (2001), in the framework of hidden Markov or semi-Markov models. These are based on plant representations motivated by biological knowledge (Barthélémy and Caraglio, 2007) and include a hidden variable, which represents plant development phases. However, such models do not account for genotypic and climatic effects.. This PhD proposal consists in extending the work by Guédon *et al.* (2001) by introducing fixed and random effects in the different distribution that define hidden semi-Markov models. Then, the aim is to propose inference algorithm with adequate accounting for model complexity. The latter is due to two levels of latent variables: hidden states corresponding to development phases ( categorical variables) and latent variables associated with random effects (continuous variables). ##### Methods and expected results Several methods are considered for inference. The first one relies on quadrature methods, whose specification and validity require a detailed study of the functions to be integrated, as in INLA (Rue *et al.*, 2009). The second one is Monte-Carlo approximation. The third one consists in Variational EM approximations (Jordan *et al.*, 1999, with VBEM as a variant in Bayesian analysis), based on an approximation of the log-likelihood with parametric functions being factorised with respect to some random variables. The work to be achieved includes a comparison of these methods regarding their abilities to accurately estimate parameters and to achieve the best possible compromise between accuracy and computation time. The issue of model selection wit usual criteria will have to be addressed, so as to determine which relevant covariates deserve to be part of the model. The selected methods for model inference will be applied on one or several available data sets in cereal root architecture, issued from research projects by Bertrand Muller [LEPSE](https://www6.montpellier.inrae.fr/lepse). Other potential applications are in fruit trees, with data issued from various agronomic contexts and projects from the [AFEF team](https://umr-agap.cirad.fr/nos-recherches/equipes-scientifiques/architecture-et-floraison-des-especes-fruitieres/liste-des-agents) (AGAP laboratory, INRAE). ##### Supervision The PhD will be co-supervised by Jean-Baptiste Durand (computational statistics), [AMAP](https://amap.cirad.fr/) Laboratory, CIRAD, and Bertrand Muller (ecophysiology), [LEPSE](https://www6.montpellier.inrae.fr/lepse). Nathalie Peyrard (computational statistics) at [MIAT research](https://miat.inra.fr/site/SCIDYN(english)) unit by INRAE Toulouse, Sandra Plancade (statistics and modelling) at MIAT and Evelyne Costes (genetics, botany and biostatistics) at AFEF team (AGAP laboratory, INRAE, Montpellier) will also take part to supervision. ##### Location and funding The PhD will take place at AMAP Laboratory, Cirad, in Montpellier (Agropolis district) https://amap.cirad.fr/. Visits to MIAT, INRAE Toulouse, will be planned: https://miat.inrae.fr/ Moreover, co-supervisers belong to [INCA](https://groupes.renater.fr/wiki/hsmm-inca/public/index) consortium (headed by N. Peyrard), which gathers most statisticians working on hidden semi-Markov models (both from theoretical and applied aspects) in France. Close collaborations, through regular working groups, will offer opportunities to interact with the community in statistics. Applicants are requested to join their full grade transcripts in third year of Bachelor, first year of Master and first semester, second year of Master (and even second semester if available), together with a resume and motivation letter. ##### Valorisation objectives The obtained results will lead to publications in computational statistics journals, in agronomy journals, and in journals in mathematical modelling for biology. The developed algorithm will be included in software packages for the statistical community in hidden semi-Markov models and adapted to the community in plant structure modelling. The PhD student will have the possibity to present his or her work in national and international workshops. ##### International collaborations The application to fruit specices (apple tree) is part of a collaboration with Martin Mészáros (Research and Breeding Institute of Pomology in Holovousy - VŠÚO - Czech Republic). An international workshop on Markovian and semi-Markovian models will be co-organised by INCA consortium in 2024. This will be an opportunity to meet international researchers with related topics and may lead to new collaborations. ##### References * Barthélémy, D. and Caraglio, Y. Plant Architecture : A Dynamic, Multilevel and Comprehensive Approach to Plant Form, Structure and Ontogeny. *Annals of Botany* **99**(3), 375–407 (2007) * Guédon, Y., Barthélémy, D., Caraglio, Y. and Costes, E. Pattern Analysis in Branching and Axillary Flowering Sequences. *Journal of theoretical biology*, **212**(4), 481–520 (2001) * Jordan, M.I., Ghahramani, Z., Jaakkola, T.S., Saul, L.K. An Introduction to Variational Methods for Graphical Models. *Machine Learning* **37**, 183–233 (1999) * Rue, H., Martino, S. and Chopin, N., Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. *Journal of the Royal Statistical Society: Series B (Statistical Methodology*), **71**: 319-392 (2009) * Yu, S.-Z. Hidden semi-Markov models. *Artificial intelligence*, **174(2)**, 215–243 (2010) -->

Publications

964018
Image document

Détection de motifs disruptifs au sein de plantes : une approche de quotientement/classification d'arborescences

Pierre Fernique , Jean-Baptiste Durand , Yann Guédon
47èmes Journées de Statistique, Société Française de Statistique, Jun 2015, Lille, France
Communication dans un congrès hal-01240305v1