Hugues Van Assel

Postdoctoral Fellow at Genentech · South San Francisco, CA

hugues-van-assel.jpeg

I am a postdoctoral fellow at Genentech with Aviv Regev, working on machine learning for science and drug discovery. My research interests include representation learning, multimodal learning, dimensionality reduction, and optimal transport.

Open source

I enjoy building and sharing research software.

TorchDR

A PyTorch toolbox for dimensionality reduction with state-of-the-art performance on single and multiple GPUs.

stable-pretraining

A PyTorch library for foundation-model pretraining with real-time training monitoring.

Background

I completed my PhD in the mathematics department at ENS Lyon, supervised by Titouan Vayer and Aurélien Garivier. During my PhD, I visited the GalilAI group at Brown University to work on self-supervised learning. Before my PhD, I studied at École Polytechnique and in the MVA master’s program.

latest posts

Nov 20, 2025 Joint-Embedding vs Reconstruction: When Should You Use Each?
Feb 25, 2024 Inverse optimal transport does not require unrolling

selected publications

  1. Joint-Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self-Supervised Learning
    Hugues Van Assel, Mark Ibrahim, Tommaso Biancalani, Aviv Regev, and Randall Balestriero
    Advances in Neural Information Processing Systems (Spotlight), 2025
    arXiv HTML Blog Poster
  2. Distributional Reduction: Unifying Dimensionality Reduction and Clustering with Gromov-Wasserstein
    Hugues Van Assel, Cédric Vincent-Cuaz, Nicolas Courty, Rémi Flamary, Pascal Frossard, and Titouan Vayer
    Transactions on Machine Learning Research, 2024
    arXiv
  3. SNEkhorn: Dimension Reduction with Symmetric Entropic Affinities
    Hugues Van Assel, Titouan Vayer, Rémi Flamary, and Nicolas Courty
    Advances in Neural Information Processing Systems, 2023
    arXiv HTML Poster
  4. A Probabilistic Graph Coupling View of Dimension Reduction
    Hugues Van Assel, Thibault Espinasse, Julien Chiquet, and Franck Picard
    Advances in Neural Information Processing Systems, 2022
    arXiv HTML Poster