- Starting date: as soon as possible
- Type: postdoctoral contract at Ecole Polytechnique.
- Contract duration: initially 2 years, with possible renewal.
- Situation: Workplace will be split between Ecole Polytechnique and College de France. You will work under the joint supervision of Filippo Vicentini, Antoine Georges and Michel Ferrero. Interactions and visits at Flatiron/CCQ will be likely.
Description
As part of a joint project AI for Correlated Materials we (Filippo Vicentini, Antoine Georges and Michel Ferrero) are hiring a Postdoctoral researcher.
The project has as an objective the development of surrogate ML models that learn the output of impurity solvers, such as the spectral function, given its input. The final step of the project is the integration of those surrogate models into a DMFT pipeline and applications to strongly correlated materials.
The proof of concept has already been realised by collaborators at CCQ, and the succesful applicant will be working with them to push the project further by working on any of the following:
- Dataset generation, which requires knowledge of existing impurity solvers
- Architecture developments, to identify the best NN architecture to learn this kind of data
- Understanding the best loss/dataset setting, and how to best frame the problem
- Collaboration with mathematicians working on formalising the problem and deriving rigorous convergence bounds
- Applications
Requirements:
- PhD in Computational Physics or Computer Science, or other related field;
- Good knowledge of Dynamical Mean Field Theory;
- Good knowledge of Machine Learning methods
- Very good coding skills
- High communication skills;