Jens Behrmann

Jens Behrmann

Postdoctoral Researcher

Apple AI/ML research, Health AI

About Me

I am a postdoctoral researcher on the Health AI team at Apple AI/ML research. Previously I was a postdoc and a PhD student at the University of Bremen, Germany. During my Ph.D. I worked on principled ways to construct neural network architectures. I am fascinated by the mathematical foundations of deep learning, in particular in generative modeling and robust invertible neural networks.

My goal is to combine a deep understanding of machine learning with an application-driven mindset. I enjoy working on problems arising in life science, see e.g. my publications on Imaging Mass Spectrometry, and broadly in industrial applications. My background is in applied mathematics, with a focus on modeling complex phenomena, signal processing, and machine learning.

Download my CV.

Interests

  • Machine Learning
  • Neural networks
  • AI/ML in life science and industry

Education

  • PhD in Deep Learning, 2019

    University of Bremen, Germany

  • M.Sc. in Industrial Mathematics, 2015

    University of Bremen, Germany

  • B.Sc. in Industrial Mathematics, 2012

    University of Bremen, Germany

Publications

Jens Behrmann, Paul Vicol, Kuan-Chieh Wang, Roger Grosse, Joern-Henrik Jacobsen (2020). Understanding and Mitigating Exploding Inverses in Invertible Neural Networks. arXiv preprint.

PDF Code

Florian Tramer, Jens Behrmann, Nicholas Carlini, Nicolas Papernot, Joern-Henrik Jacobsen (2020). Fundamental Tradeoffs between Invariance and Sensitivity to Adversarial Perturbations. In ICML.

PDF Code

Alexander Denker, Maximilian Schmidt, Johannes Leuschner, Peter Maass, Jens Behrmann (2020). Conditional Normalizing Flows for Low-Dose Computed Tomography Image Reconstruction. In ICML INNF+ workshop.

PDF

Ricky T. Q. Chen, Jens Behrmann, David Duvenaud, Joern-Henrik Jacobsen (2019). Residual Flows for Invertible Generative Modeling. In NeurIPS.

PDF Code

Christian Etmann, Maximilian Schmidt, Jens Behrmann, Tobias Boskamp, Lena Hauberg-Lotte, Annette Peter, Rita Casadonte, Joerg Kriegsmann, Peter Maass (2019). Deep Relevance Regularization: Interpretable and Robust Tumor Typing of Imaging Mass Spectrometry Data. arXiv preprint.

PDF Project

See all publications