Seyed Mohsen Moosavi-Dezfooli

Research Scientist

I am a Staff Machine Learning Researcher at Apple Zurich. Prior to that, I was an Assistant Professor at Imperial College London in the Department of Electrical and Electronic Engineering. Before joining Imperial College, I was a postdoctoral researcher in the Institute for Machine Learning at the ETH Zurich. I received my PhD in Computer and Communication Sciences under the supervision of Pascal Frossard at the École Polytechnique Fédérale de Lausanne (EPFL), Switzerland in November 2019. My PhD research focused on the analysis of the robustness of classifiers and its connection with the geometry of deep neural networks. I was also fortunate to intern for the Apple AI Research in 2017, working on autonomous technologies.

I got my M.Sc. degree in Communication Systems from the EPFL, in 2014, and my B.Sc. in Electrical Engineering from Tehran Polytechnic, Iran, in 2012.

News

Jun 2024 I have joined Apple Zurich as an AI/ML researcher!
Aug 2022 I gave an invited talk on various aspects of robustness and generalization at Apple MLR/MI.
Jul 2022 I attended Dagstuhl Seminar on Security of Machine Learning. That was amazing!
Jun 2022 Our work on data augmentationt for common corruptions, PRIME, got accepted at ECCV 2022!
Mar 2022 My joint work with EPFL’s VITA lab on robustness of vehicle trajectory prediction models is accepted at CVPR 2022.

Selected Publications

  1. ICLR
    Generic placeholder image
    Reducing Excessive Margin to Achieve a Better Accuracy vs. Robustness Trade-off
    Rahul Rade, and Seyed-Mohsen Moosavi-Dezfooli
    In International Conference on Learning Representations 2022
    PDF
  2. ICML
    Generic placeholder image
    Uniform Convergence, Adversarial Spheres and a Simple Remedy
    Gregor Bachmann, Seyed-Mohsen Moosavi-Dezfooli, and Thomas Hofmann
    In International Conference on Machine Learning 2021
    arXiv
  3. NeurIPS
    Generic placeholder image
    Neural Anisotropy Directions
    Guillermo Ortiz-Jimenez, Apostolos Modas, Seyed-Mohsen Moosavi-Dezfooli, and Pascal Frossard
    In Advances in Neural Information Processing Systems 2020
    arXiv Blog Code Video
  4. CVPR
    Generic placeholder image
    GeoDA: a geometric framework for black-box adversarial attacks
    Ali Rahmati, Seyed-Mohsen Moosavi-Dezfooli, Pascal Frossard, and Huaiyu Dai
    In IEEE Conference on Computer Vision and Pattern Recognition 2020
    arXiv Code Video
  5. CVPR
    Generic placeholder image
    Robustness via curvature regularization, and vice versa
    Seyed-Mohsen Moosavi-Dezfooli, Alhussein Fawzi, Jonathan Uesato, and Pascal Frossard
    In IEEE Conference on Computer Vision and Pattern Recognition 2019
    arXiv Code
  6. CVPR
    Generic placeholder image
    Geometric robustness of deep networks: analysis and improvement
    Can Kanbak, Seyed-Mohsen Moosavi-Dezfooli, and Pascal Frossard
    In IEEE Conference on Computer Vision and Pattern Recognition 2018
    arXiv Code