ETH4D fellowship for the project at ETH Zurich
I have been awarded ETH4D fellowship to support my research stay at ETH Zurich, Switzerland. I will be working on mathematical analysis of Neural Operators and their applications in computational PDEs with Prof. Dr. Siddhartha Mishra at CAMLab. Specifically, I worked on proving universal approximation property (UAP) for Geometry aware Operator transformer (GAOT) using techniques from classical PDE analysis, functional and Fourier analysis.
The period of my stay is May ~ August, 2025.
[Edit: Aug, 2025] Universal approximation property for GAOT states that: GAOT can approximate any continuous solution operator between \(L^p\) function spaces to any arbitrary accuracy. I was able to show that GAOT is indeed a universal approximator and hence providing a theoretical foundation to the model.
