PhD
Theoretical Foundations of Artificial Intelligence
Technical University of Munich
Faculty of Informatics – I7
Boltzmannstr. 3
85748 Garching bei München
Biosketch
Maedeh obtained both her Bachelor’s and Master’s in Computer Science. She did her master’s in Heidelberg in collaboration with EMBL. Her research interests are explainability, theory of deep learning, and Kernel methods. Maedeh is currently focusing on different approaches towards interpretable Machine Learning models as a PhD student under supervision of Prof. Debarghya Ghoshdastidar at TUM.
relAI Research
Interpretable and Generative Modeling
I previously worked on interpretability in self-supervised learning and foundation models via an SSL-suited formulation of influence functions leveraging neural tangent kernel surrogate models. Currently, I’m focusing on the theory of Wasserstein-Fisher-Rao gradient flow to formulate the problem of missing value imputation in the missing at random (MAR) case, as an unbalanced optimal transport problem. Moreover, I will provide consistency and convergence guarantees for the proposed formulation.