Hi, I’m Johann. I’m currently a PhD student in the math department at UC San Diego. Some true statements about my interests/research:
- I like to combine elegant theory with useful practice.
- I want to find and understand the natural (``right’’) way to do things in deep learning.
- In particular I like to understand the geometry and dynamics of (deep) learning related things.
- For example: How should the optimizer look like, or generally how should the network be trained? How should hyperparameters of network architecture and optimizer be chosen? I think, to some degree, there is natural/canonical answer to these questions.
- My recent work has been in quantization/compression of neural networks.
Feel free to contact me via [firstname].[lastname]@hotmail.de, I’m always happy to speak about math, cs, physics, engineering, sports, entrepreneurship, life, and everything else.
This subpage was last updated in October 2026.