Faculty
Dr. Rebekka Burkholz is a tenured faculty at the CISPA Helmholtz Center for Information Security in Saarbrücken. The focus of her research is relational machine learning. Her main goal is to gain a theoretical understanding of deep learning from a complex network perspective and improve contemporary algorithms based on these insights. Her favourite applications are currently in molecular biology.
From 2019-2021, she was a PostDoc at the Biostatistics Department of the Harvard T.H. Chan School of Public Health working with John Quackenbush. Before that, she enjoyed postdoctoral research at ETH Zurich, from 2017-2018 at the Institute for Machine Learning with Joachim Buhmann and from 2016-2017 at the Chair of Systems Design with Frank Schweitzer. Her PhD research from 2013-2016 at the ETH Risk Center was supervised by Frank Schweitzer and co-supervised by Hans J. Herrmann. Her thesis on systemic risk won the Zurich Dissertation Prize and her work on international maize trade received the CSF Best Contribution Award. She studied Mathematics and Physics at TU Darmstadt.
International Conference on Machine Learning (ICML)
Mirror, Mirror of the Flow: How Does Regularization Shape Implicit Bias?
Conference on Neural Information Processing Systems (NeurIPS)
International Conference on Learning Representations (ICLR)
GNNs Getting ComFy: Community and Feature Similarity Guided Rewiring
International Conference on Learning Representations (ICLR)
Mask in the Mirror: Implicit Sparsification
Conference on Neural Information Processing Systems (NeurIPS)
Pruning neural network models for gene regulatory dynamics using data and domain knowledge
Conference on Neural Information Processing Systems (NeurIPS)
Spectral Graph Pruning Against Over-Squashing and Over-Smoothing
Conference on Neural Information Processing Systems (NeurIPS)
Dynamic Rescaling for Training GNNs
International Conference on Machine Learning (ICML)
GATE: How to Keep Out Intrusive Neighbors
Genome Biology Biologically informed NeuralODEs for genome-wide regulatory dynamics
International Conference on Learning Representations (ICLR)
Batch normalization is sufficient for universal function approximation in CNNs.