Antoine CARADOT

Developed advanced ML algorithms for physics-based modeling, focusing on adaptive sampling and uncertainty quantification in PINNs.

Developed methods to discover governing equations (PDEs) from sparse, noisy data, improving how physical laws can be inferred when observations are limited.

Published at ECML PKDD 2025 and GRETSI 2025 on provably accurate and stable learning frameworks for sparse data problems.

Recherche en théorie des représentations avec des chercheurs locaux. Participation aux séminaires organisés sur le campus.

Generalised known algebraic structures used in string theory to a larger context, opening the way to new descriptions in both algebra and geometry.

Developed a method of construction of a type of algebraic structure (differential graded vertex algebra) from a simpler, more easily manipulated object (differential graded vertex Lie algebra), offering a way to obtain detailed and meaningful examples.

Stayed in the USA (Kansas State University) for four months to collaborate with colleagues on a generalisation of the cohomological variety (see below) to the theory of modules, by analogy with known work on group theory and finite dimensional algebra theory.

Published original research articles in international mathematical journals.

Introduced and investigated a new “cohomological variety” concept for vertex algebras, defining and exploring a dual structure to the traditional “associated variety”, thereby contributing deeper insight into the internal symmetries and invariants of algebraic systems.

Studied the invariants of a type of algebraic structure (non rational vertex algebra) and devised a characterisation of the simplicity of its internal structure using category theory and homological algebra.

Published original research articles in international mathematical journals.