Violaine Courrier

I am Violaine Courrier, a PhD student in the Datavers team at INRIA Lille working in an industrial partnership with the French company Withings. Under the supervision of Christophe Biernacki and Cristian Preda, my research focuses on deep unsupervised learning, specifically addressing the challenge of identifying relevant clusters within complex data, such as connected health time series.

To tackle this, my work formalizes the paradigm of “guided clustering”. It uses an external contextual variable to steer the algorithm toward the partition most relevant for a given task. I develop deep generative models, including the Guided Clustering Variational Autoencoder (GCVAE) , and I am currently working on its dynamic extensions for time series (DGCVAE). Through my work at Withings, my goal is to apply these methods to extract meaningful physiological and behavioral profiles from multivariate connected health data.

My research interests include:

• Clustering & unsupervised learning: identifying meaningful partitions within complex data.
• Deep generative models: flexible probabilistic architectures to learn structured representations.
• Statistical learning theory: theoretical guarantees and statistical consistency for generative models.
• Sequential data modeling: capturing temporal dynamics in multivariate time series.
• Applied ML for real-world health data: extracting meaningful physiological and behavioral profiles from complex, real-world connected health data to drive healthcare applications.