Olivier J. HénaffSenior Staff Research Scientist Google DeepMind, London, UK |
My research aims to understand the principles underlying biological and artificial intelligence. Humans and animals produce amazingly complex behaviors, in the face of changing environments and often with little to no supervision. This has led me to study the structure of neural representations in perceptual and physiological experiments, asking how our visual system might enable such behaviors.
At DeepMind, I have been investigating self-supervised algorithms that extract structure from raw data, enabling data-efficient image recognition, behaviorally-relevant scene understanding, and unsupervised object discovery and temporal correspondence. More recently, I have been interested in how visual representations might structure our memory, enabling flexible perceptual inference and long-video understanding.
Prior to joining DeepMind, I completed my PhD at NYU's Center for Neural Science, advised by Eero Simoncelli. Before starting research, I studied math and physics at École Polytechnique and Lycée Sainte Geneviève.
Data curation via joint example selection further accelerates multimodal learning Neural Information Processing Systems (NeurIPS), December 2024 (Spotlight) |
|
Memory Consolidation Enables Long-Context Video Understanding International Conference on Machine Learning (ICML), May 2024 (Spotlight) |
|
Bad Students Make Great Teachers: Active Learning Accelerates Large-Scale Visual Understanding European Conference on Computer Vision (ECCV), February 2024 |
|
Towards In-context Scene Understanding Neural Information Processing Systems (NeurIPS), December 2023 (Spotlight) |
|
Self-supervised video pretraining yields human-aligned visual representations Neural Information Processing Systems (NeurIPS), December 2023 |
|
Object discovery and representation networks European Conference on Computer Vision (ECCV), October 2022 |
|
Primary visual cortex straightens natural video trajectories Nature Communications, October 2021. *equal contribution |
|
Efficient Visual Pretraining with Contrastive Detection International Conference on Computer Vision (ICCV), October 2021 (Oral) |
|
Divide and Contrast: Self-supervised Learning from Uncurated Data International Conference on Computer Vision (ICCV), October 2021 |
|
Data-Efficient Image Recognition with Contrastive Predictive Coding International Conference on Machine Learning (ICML), July 2020 |
|
Are we done with ImageNet? Tech report, June 2020. *equal contribution |
|
Representation of visual uncertainty through neural gain variability Nature Communications, May 2020 |
|
Perceptual straightening of natural videos Nature Neuroscience, April 2019 |
|
Geodesics of learned representations International Conference on Learning Representations (ICLR), May 2016 |
|
The local low-dimensionality of natural images International Conference on Learning Representations (ICLR), May 2015 (Oral) |