Short bio

From January 2021 to October 2024, I was a PhD Student in the IECL lab at Université de Lorraine (Nancy, France). I am focusing on Statistics and Machine Learning, and especially time series analysis. My advisors are Pr. Marianne Clausel and Pr. Konstantin Usevich. Before that, I earned a Bachelor’s degree in Mathematics and a Master’s degree in Data Science both from Sorbonne Université (Paris, France). In addition, through internships, I have been practicing Data Science at leading companies and research institutes.

Research interests

My thesis focused on the use of the signature, a tool from rough paths theory, for time series analysis and statistical learning. This tool extract intrinsic information about multidimensional time series and more precisely the dependencies in variations along the space dimensions. Find here an introduction to this tool.

News

October 2024: 🎉 PhD defended successfully ! The manuscript is available in English at https://theses.fr/2024LORR0165 (pdf here).
April 2024: 🤓 Workshop “Mathematics of data streams: signatures, neural differential equations, and diffusion models” (website) in Greifswald, Germany !

Publications

R. Mignot, V. Mange, K. Usevich, M. Clausel, J.-Y. Tourneret and F. Vincent, “Anomaly Detection Using Multiscale Signatures”. In: Proceedings of the 32nd European Signal Processing Conference (EUSIPCO), 2024, pp.2757-2761. DOI: 10.23919/EUSIPCO63174.2024.10714963. PDF. Software.
J. Cugliari, E. Devijver, A. Meynaoui, and R. Mignot, “Some recent developments on functional data analysis”. In: ESAIM: Proceedings and Surveys, EDP Sciences, 2024, to appear.
R. Mignot, M. Clausel, K. Usevich and N. Sugiura, “Principal geodesic analysis for time series encoded with signature features.” Preprint hal-04392568, 2024. Software.
M. Clausel, J. Diehl, R. Mignot, L. Schmitz, N. Sugiura, and K. Usevich, “The barycenter in free nilpotent Lie groups and its application to iterated-integrals signatures”. In: SIAM Journal on Applied Algebra and Geometry 8.3 (2024), pp.519-552. DOI: 10.1137/23M159024X. Software.

Talks

2024/10 PhD thesis defense: University of Lorraine, Nancy, France
2024/08 Conference: EUSIPCO 2024 (website). Lyon, France
2024/04 Workshop: Mathematics of data streams. (website). Greifswald, Germany
2023/12 Seminar: Signature, applications and Machine Learning (website). University of Pau and the Adour Region, Pau, France
2023/11 Group seminar. University of British Columbia, Vancouver, Canada
2023/11 Probability & Stat. Seminar (website) University of Luxembourg, Esch-sur-Alzette, Luxembourg
2023/10 Workshop: Mathematical Founding Principles of AI (website) Sorbonne University, Paris, France
2023/09 Lorraine - Luxembourg Workshop in Probability and Statistics (website). University of Lorraine & University of Luxembourg. Metz, France
2023/07 Group seminar (website). University Lyon 2. ERIC Lab. Lyon, France
2023/06 Workshop GDR TRAG (Research group on Rough Paths) (website). Paris-Dauphine University, Paris, France
2023/01 Group seminar. University Grenoble–Alpes, Computer Sci. Lab (LIG). Grenoble, France
2022/09 Conference: 28th GRETSI (Signal and Image Processing) (website). Session “Statistical learning on Lie groups”. Nancy, France
2022/08 Conference: MAS (Modeling, Stochastic and Statistics) (website). Rouen, France
2022/07 Group seminar. Japan Agency for Marine-Earth Science and Technology. Tokyo, Japan
2022/06 Conference: French Statistics Society annual conference (website). Lyon, France

Teaching

2022 - 2024 : Data Analysis (Tutorials, 25h each year). Ecole des Mines de Nancy. Level: Graduate. Topics: R, Linear models, Principal Component Analysis, Factor analysis, Classification, Linear discriminant analysis, Logistic regression, Association rule learning.
2021 - 2024 : Probability and Statistics (Tutorials, 20h each year). University of Lorraine. Level: Graduate. Topics: Point estimation, Confidence intervals, Hypothesis testing (parametric tests, two-sample tests, chi-square tests), Python.

Softwares

See Software links in the publication section above 👆.