Konstantin Schürholt

Konstantin Schürholt

Machine Learning researcher investigating populations of Neural Networks, self-supervised and meta learning, sparsity and ML for science.

Publications

Sparsified Model Zoo Twins: Investigating Populations of Sparsified Neural Network Models

Published in ICLR Workshop on Sparsity in Neural Networks 2023, 2023

In this paper, we apply common sparsification methods on population of Neural Networks and analyze their performance and relation between sparse and full models.

Recommended citation: Honegger et al., 2023. "Sparsified Model Zoo Twins: Investigating Populations of Sparsified Neural Network Models." ICLR Workshop on Sparsity in Neural Networks 2023. https://arxiv.org/abs/2304.13718

Towards dynamic stability analysis of sustainable power grids using graph neural networks.

Published in NeurIPS 2022 Climate Change AI Workshop, 2022

We generate new datasets based on dynamical simulations of power grids as a challenge for Graph Neural Networks and include benchmark performances on different tasks including out-of-distribution generalization.

Recommended citation: Nauck et al., 2022. "Towards dynamical stability analysis of sustainable power grids using Graph Neural Networks." NeurIPS 2022 Workshop on Tackling Climate Change with Machine Learning 2022. https://www.climatechange.ai/papers/neurips2022/16

Hyper-Representation for Pre-Training and Transfer Learning

Published in International Conference on Machine Learning (ICML) Pretraining Workshop 2022, 2022

In this paper, we extend hyper-representations for generative use to sample neural network weights for initialization and transfer learning.

Recommended citation: Schürholt et al., 2022. "Hyper-Representation for Pre-Training and Transfer Learning." ICML Pretraining Workshop 2022. https://arxiv.org/abs/2207.10951

Self-Supervised Representation Learning on Neural Network Weights for Model Characteristic Prediction

Published in Neural Information Processing Systems (NeurIPS) 2021, 2022

This paper proposes to learn self-supervised representations of the weights of populations of NN models using novel data augmentations and an adapted transformer architecture.

Recommended citation: Schürholt et al., 2021. "Self-Supervised Representation Learning on Neural Network Weights for Model Characteristic Prediction." NeurIPS 2021. https://arxiv.org/abs/2110.15288

Elements of future snowpack modeling – Part 1: A physical instability arising from the nonlinear coupling of transport and phase changes

Published in The Cryosphere, 2022

This paper studies the effect of numerical instabilities of coupled heat and mass equations of ice in snow pack.

Recommended citation: Schürholt et al., 2022. "Elements of future snowpack modeling – Part 1: A physical instability arising from the nonlinear coupling of transport and phase changes." The Cryosphere 2022. https://tc.copernicus.org/articles/16/903/2022/

On water vapor transport in snowpack models: Comparison of existing schemes, numerical requirements and the role of non-local advection.

Published in European Geosciences Union (EGU) General Assembly 2019, 2019

In this work, we investigate the numerical properties of coupled mass and energy equations of snow pack.

Recommended citation: Schürholt et al., 2019. "On water vapor transport in snowpack models: Comparison of existing schemes, numerical requirements and the role of non-local advection." EGU General Assembly 2019. https://search.ebscohost.com/login.aspx?direct=true&profile=ehost&scope=site&authtype=crawler&jrnl=10297006&AN=140480720&h=OIy4BTnTfn8MswNyU913MD0Xo04OcI6gH7aYV7TAye0vFuoL6%2FrVJwyn3PeyJIlnbWBhd97x9Iezqwkcgn5k0w%3D%3D&crl=c