Maren Mahsereci

I am a postdoctoral researcher at the University of Tübingen. Previously I have been at Amazon Web Services (AWS) Berlin and Amazon Retail (AMZN) Cambridge as Machine Learning Scientist. I graduated from the Max-Planck Institute for Intelligent Systems (MPI) Tübingen.

My research interests are stochastic optimization methods for deep learning, probabilistic numerics, and statistical emulation. I am also an active contributor and maintainer of open source software projects on GitHub.

news

Mar 28, 2022 I gave a talk about the open source library ProbNum at the Workshop on Probabilistic Approaches to Numerical Computation in London, UK.
Dec 3, 2021 The paper on our open source project ProbNum: Probabilistic Numerics in Python is now on arXiv. Please cite if you use ProbNum in your scientific work (ProbNum webpage).
Oct 23, 2021 Our paper Invariant Priors for Bayesian Quadrature has been accepted as a contributed talk at the NeurIPS 2021 workshop Your Model is Wrong: Robustness and misspecification in probabilistic modeling.
Oct 1, 2021 I am co-organizing the Dagstuhl Seminar on Probabilistic Numerical Methods - From Theory to Implementation Wadern, Germany.
Jul 31, 2021 Our paper Dynamic Pruning of a Neural Network via Gradient Signal-to-Noise Ratio of my intern Julien Siems at Amazon Web Services has been accepted at the 8th ICML Workshop on Automated Machine Learning (AutoML) 2021.