Tensor-Train Methods for Sequential State and Parameter Learning in State-Space Models

In this thesis, we tackle real-world, model-constrained learning challenges involving dynamic systems and incomplete data. The goal is to predict and control hidden behaviours of dynamical systems. Our research introduces innovative methods using tensor-train techniques to solve this problem. Using our methods, one can explain complex systems and predict their possible evolution. In simpler terms, we are developing smart ways to understand and manipulate real-world systems even when we do not have all the pieces of the puzzle.