Thesis defense Maxime Metz

Thesis defense: Maxime Metz

26 November 2021

Agropolis International, amphitheatre "Louis Malassis"

Maxime will defend his PhD thesis #DigitAg in partnership with LIRMM and CIRAD, entitled "Development of chemometric methods for massive data processing"

Friday November 26th 9:30am, Maxime Metz will defend his PhD thesis:

Development of chemometric methods for massive data processing

Chemical data analysis, also known as chemometrics, is widely used in agronomy to address various issues such as soil, forage or phenotyping studies. Today, a large amount of data can be generated and chemometricians must be able to analyse them. However, their usual tools are not able to process this amount of data efficiently. Tools originated from big data field have been developed to process large databases but have not yet been evaluated for chemometrics. The objective of this thesis is therefore to study massive data processing for chemometrics. To this end, three research axis were investigated. The first research axis is to study how to enable massive data processing by local methods. Local methods calibrate a model per individual to be predicted on its nearest neighbours. The second research axis is to study the relevance of an individual within a local model. The third research axis consists in combining the ideas developed in the first two axis to propose efficient methods for chemometrics.To address the first research focus, a new method called parSketch-PLS was studied and developed. To address the second focus, a method called RoBoost-PLSR was developed. To address the third focus, two premise methods were proposed. The results associated with these developments highlighted the interest in adapting massive data processing tools to chemometrics. Indeed, tools used for massive data processing do not necessarily rely on the same knowledge as tools developed for chemometrics. This can degrade the predictive capacity of chemometrics.
This thesis therefore highlights the interest in bringing these two fields together in order to propose a set of methods and tools for processing massive chemical data. 

 

The jury is composed of the following members:

  • Gilbert Saporta, Professor Emeritus, CNAM, France
  • Douglas Rutledge, Professor Emeritus, AgroParisTech, France
  • Florent Masseglia, Senior Researcher, INRIA,France
  • Fédérico Marini, Profesor, Université de Rome, Italie
  • Marina Cocchi, Associate Professor, Université de Modène et de Reggio d'Émilie, Italie
  • Jean-Michel Roger, Research Engineer, INRAE-ITAP, France
  • Matthieu Lesnoff, Researcher, CIRAD, France
  • Reza Akbarinia, Researcher, INRIA, France

Contact: changeMe@inrae.fr

Publication date : 17 July 2023