David Berthelot

Ars Facere (to make skill - an allusion to AI) is a blog about machine learning, linux, gaming, etc… Disclaimer: The opinions expressed here are my own and do not represent the ones of my employer.

About me

I initially started as a software engineer and later began doing machine learning as a hobby in my free time. Since then I’ve become a research scientist on machine learning (in a large company).

Interests

  • Semi-Supervised Learning (SSL)

  • Representation learning

  • Super-Resolution (SR)

  • Generative Adversarial Networks (GANs)

See Selected Publications for more details.

Programming wise, I’m the lead developer of Objax, a machine learning framework with a similar interface to PyTorch but made on top of JAX for performance and flexibility.

Latest posts

  • 07 October - Sequential Tinkering - Part 1

    Objax offers a significantly enhanced version of the ubiquitous Sequential layer. In this first part of a two-part series, I explore some of its basic, yet powerful, features.

    _images/lego-bridge.jpg

  • 15 April - Windows GPU gaming on Linux

    Gaming on Linux with virtual machine running Windows takes a little bit of work, specifically on how to make the GPU visible to windows. In practice, you’ll need two GPUs (one for Linux, one for the Windows virtual machine (VM)). This guide explains how make a GPU visible to the Windows VM.

    _images/witcher.jpg

Selected Publications

LAG
Creating High Resolution Images with a Latent Adversarial Generator
David Berthelot, Peyman Milanfar, Ian Goodfellow
arXiv preprint arXiv:2003.02365
FixMatch
Fixmatch: Simplifying semi-supervised learning with consistency and confidence
Kihyuk Sohn 1, David Berthelot 1, Chun-Liang Li, Zizhao Zhang, Nicholas Carlini, Ekin D Cubuk, Alex Kurakin, Han Zhang, Colin Raffel
arXiv preprint arXiv:2001.07685
ReMixMatch
Remixmatch: Semi-supervised learning with distribution alignment and augmentation anchoring
David Berthelot, Nicholas Carlini, Ekin D Cubuk, Alex Kurakin, Kihyuk Sohn, Han Zhang, Colin Raffel
International Conference on Learning Representations. 2020
MixMatch
MixMatch: A Holistic Approach to Semi-Supervised Learning
David Berthelot, Nicholas Carlini, Ian Goodfellow, Nicolas Papernot, Avital Oliver, Colin Raffel
Advances in Neural Information Processing Systems, 2019
ACAI
Understanding and Improving Interpolation in Autoencoders via an Adversarial Regularizer
David Berthelot 1, Colin Raffel 1, Aurko Roy, Ian Goodfellow
International Conference on Learning Representations. 2019
BEGAN
BEGAN: Boundary Equilibrium Generative Adversarial Networks
David Berthelot, Thomas Schumm, Luke Metz
arXiv preprint arXiv:1703.10717

Footnotes

1(1,2,3,4)

Co-first authors.