PhD Position on Generative Modeling: This PhD thesis project at IMT Atlantique focuses on developing generative models to create constrained-valued images for inverse problems in remote sensing, specifically in the context of hyperspectral images.
PhD Position on Generative Modeling for Constrained Bayesian Inverse Problems in Imaging
Designation
PhD Candidate
| Field | Details |
|---|---|
| Research Area | Generative Modeling, Bayesian Inverse Problems, Imaging |
| Location | IMT Atlantique, Brest, France |
| Eligibility/Qualification | Relevant Master’s degree in fields such as Computer Science, Mathematics, Engineering, or related areas. Experience in machine learning or imaging techniques is desirable. |
| Last Date to Apply | January 31, 2026 |
Description
The goal of the PhD project is to develop generative models that can create constrained-valued 2D images, departing from classical approaches that work well with unconstrained images. The research aims to utilize constrained Gaussian Processes as a baseline before advancing to flow-matching techniques for more realistic generative priors. The initial focus will be on the spectral unmixing problem within hyperspectral images, which requires that each pixel’s values reside within a probability simplex. Depending on the findings, additional constraints and applications may be explored in the future.
How to Apply
Interested candidates should submit their application and supporting documents as outlined in the official guidelines on the IMT Atlantique website. Please refer to the application portal linked below for detailed instructions.
This scholarship opportunity presents a unique chance for motivated individuals to engage in cutting-edge research in imaging and machine learning. Make sure to prepare a compelling application before the deadline!
