Seijin Kobayashi

I am a Machine Learning PhD student supervised by Prof. Angelika Steger at the Institute of Theoretical Computer Science, ETH Zurich. I received my bachelor in applied mathematics from École polytechnique (Palaiseau, France) in 2014 and my Master's degree from ETH Zürich in computer science in 2016, after which I worked for 3 years at Google Zurich as a Software Engineer. My interest resides at the intersection between meta learning, a specific class of neural network parametrization called hypernetworks, and reinforcement learning. I am also interested in pushing the boundaries of Deep Reinforcement Learning with a focus on computational efficiency drawing inspiration from bounded rationality in humans, as well as shedding light on inductive biases induced by deep learning models in general.

Publications

The least-control principle for learning at equilibrium
Alexander Meulemans*, Nicolas Zucchet*, Seijin Kobayashi*, Johannes von Oswald, João Sacramento,
NeurIPS, 2022. 

Learning where to learn: Gradient sparsity in meta and continual learning
Johannes Von Oswald*, Dominic Zhao*, Seijin Kobayashi, Simon Schug, Massimo Caccia, Nicolas Zucchet, João Sacramento,
NeurIPS, 2021. 

Posterior meta-replay for continual learning
Christian Henning*, Maria Cervera*, Francesco D'Angelo, Johannes Von Oswald, Regina Traber, Benjamin Ehret, Seijin Kobayashi, Benjamin F Grewe, João Sacramento,
NeurIPS, 2021. 

On the reversed bias-variance tradeoff in deep ensembles
Seijin Kobayashi*, Johannes Von Oswald*, Benjamin F. Grewe,
ICML 2021 Workshop on Uncertainty and Robustness in Deep Learning, 2021. 

Neural networks with late-phase weights
Johannes von Oswald*, Seijin Kobayashi*,Alexander Meulemans, Christian Henning, Benjamin F. Grewe, João Sacramento
ICLR, 2021. 

Meta-learning via hypernetworks
Dominic Zhao, Seijin Kobayashi, João Sacramento, Johannes von Oswald,
NeurIPS Workshop on Meta-Learning, 2020. 

Check Google scholar for an updated and complete publication history.