Efficient Uncertainty Quantification For Full Waveform Inversion Via Shot-Encoded Hessian
IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2026
I am a PhD student in computer vision at the Technical University of Munich (TUM), mentored by Prof. Daniel Cremers. My research focuses on visual perception and geometric understanding of the physical world, particularly 3D/4D reconstruction and scene understanding.
Previously, I received my Master’s degree in Mathematics from TUM, where I wrote my thesis on self-supervised Visual SLAM under Prof. Daniel Cremers. During my Master’s, I spent a semester as a visiting student at EPFL, working with Prof. Lénaïc Chizat and Prof. Laura Grigori. Before that, I completed my Bachelor’s degree in Mathematics at TUM under the supervision of Prof. Andreas S. Schulz.
Alongside my studies, I worked as a student researcher at the German Aerospace Center (DLR), focusing on autonomous exploration with robotic swarms and physics-based deep learning. Prior to that, I was a student software developer at SAP.
IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2026
Published:
This post surveys theoretical explanations of in-context learning based on Transformers implementing optimization algorithms in their forward pass. It reviews how linear attention can realize preconditioned gradient descent for linear regression and how nonlinear, kernel-based attention generalizes this mechanism to functional gradient descent, together with the underlying assumptions and limitations.