About
I work as a research scientist in the robust and verified AI team at DeepMind. Prior to this, I completed my DPhil / PhD under the supervision of Andrew Zisserman and Pawan Kumar at University of Oxford.
Research interests: optimization, deep learning, verification and privacy-preserving machine learning.
Featured Publications
Unlocking High-Accuracy Differentially Private Image Classification through Scale.
Soham De, Leonard Berrada, Jamie Hayes, Samuel L. Smith, Borja Balle. arXiv (2022).
Arxiv link
Github link
Verifying Probabilistic Specifications with Functional Lagrangians.
Leonard Berrada, Sumanth Dathathri, Krishnamurthy Dvijotham, Robert Stanforth, Rudy Bunel, Jonathan Uesato, Sven Gowal, M. Pawan Kumar. NeurIPS (2021).
Arxiv link
Github link
Training Neural Networks for and by Interpolation.
Leonard Berrada, Andrew Zisserman, M. Pawan Kumar. ICML (2020).
Arxiv link
Github link
Smooth Loss Functions for Deep Top-k Classification.
Leonard Berrada, Andrew Zisserman, M. Pawan Kumar. ICLR (2018).
Arxiv link
Github link
News
- Our paper on verification of probabilistic specifications of neural networks has been accepted as a spotlight at NeurIPS 2021!
- I’m glad to have been rated in the top 10% reviewers at NeurIPS 2020.
- I joined the Robust and Verified AI team at DeepMind as a research scientist (March 2020).
- My thesis was accepted without correction by my examiners Andrea Vedaldi and Julien Mairal (25 February 2020). Many thanks to Yougov and the EPSRC for having funded my PhD!
- I did an internship at DeepMind in the team of James Martens (June 2019 - November 2019).
- I am honored to be acknowledged in Gilbert Strang’s new book: Linear Algebra and Learning From Data.
- I gave a talk at the Machine Learning Meetup of London (August 2018).
- I presented my work at the International Symposium on Mathematical Programming (July 2018).
- Outstanding Reviewer Award at CVPR 2018.