Postdoc position in Mathematics of Machine Learning

A postdoctoral position is available to work with Eddie Aamari on mathematical and statistical aspects of modern machine learning, in connection with the PEPR IA project PERSNET.

The successful candidate will be based full-time in Paris, within the Probability and Statistics team of the DMA, ENS-PSL, and at the ENS-PSL Center for Data Science.

Possible topics

• Topological and geometrical analysis of neural networks.
• Transformers, attention dynamics, tokenization and learned representations.
• Generative modeling on manifolds, stratified spaces and singular data.
• Diffusion and score-based models, curvature, reach and tangent/normal geometry.
• Drifting models, kernelized transport, MMD/RKHS gradient flows and nonlocal PDEs.
• RKHS geometry, kernel mean embeddings and geometric statistical testing.

Profile

We seek a candidate with a strong mathematical background in probability, statistics, geometric inference, optimal transport, PDEs, stochastic processes, TDA, kernel methods, or mathematical aspects of machine learning.

Experience with generative models, transformers or numerical experiments is welcome but not required. The main criterion is mathematical strength and independence.

Application

Applications should be sent to eddie.aamari@ens.fr with subject line Postdoc application - PERSNET.

Please include:

• a CV with publications and preprints;
• a short motivation letter;
• one or two representative papers;
• the names and contacts of two or three academic references.