PeopleofSuperintelligence

Researcher

Pascal Vincent

Pascal Vincent is a deep-learning researcher whose denoising autoencoders were a precursor of self-supervised learning and whose denoising score matching principle is now routinely used to train diffusion-based generative models. He earned a PhD at the Université de Montréal under Yoshua Bengio in 2003 and coauthored the 2003 paper "A Neural Probabilistic Language Model". His denoising autoencoder papers appeared in 2008 and 2010, a 2011 paper tied score matching to them, and his work on regularizing autoencoders was at the heart of the strategy that won the 2011 Unsupervised and Transfer Learning Challenge. He joined Facebook AI Research in Montreal in 2017 and became a full-time research scientist there in 2020. He is a research scientist in the FAIR team at Meta and an adjunct professor at the Université de Montréal.

Badges

Timeline

  1. 2003

    Doctorate under Yoshua Bengio

    He earns a PhD at the Université de Montréal in 2003 with the thesis "Modèles à noyaux à structure locale". Yoshua Bengio is his doctoral advisor.

  2. 2003

    A neural probabilistic language model

    In 2003, he coauthors "A Neural Probabilistic Language Model" with Yoshua Bengio, Réjean Ducharme and Christian Jauvin. The model learns a distributed representation for each word together with the probability function for word sequences.

  3. 2008–2010

    Denoising autoencoders for deep networks

    His denoising autoencoder work of 2008 and 2010 is the first to propose filling in artificially introduced blanks to learn useful representations in any modality, a precursor of self-supervised learning. The 2010 paper, written with Hugo Larochelle, Isabelle Lajoie, Yoshua Bengio and Pierre-Antoine Manzagol, stacks layers of denoising autoencoders trained locally to denoise corrupted inputs, and the resulting deep networks narrow the gap with deep belief networks and in several cases surpass them.

  4. 2011

    Denoising score matching

    In 2011, he publishes "A Connection Between Score Matching and Denoising Autoencoders" and develops the denoising score matching principle. Diffusion-based generative models now use it routinely for training.

  5. 2011–2012

    Challenge win and student awards

    His work on regularizing denoising and contractive autoencoders is at the heart of the strategy that wins the 2011 Unsupervised and Transfer Learning Challenge. He receives an Outstanding Student Paper honorable mention at NIPS in 2011 and a Best Student Paper Award at ICML in 2012.

  6. 2012–2013

    A review of representation learning

    In 2012, he coauthors "Representation Learning: A Review and New Perspectives" with Yoshua Bengio and Aaron Courville, a review of unsupervised feature learning and deep learning that covers probabilistic models, auto-encoders, manifold learning and deep networks. It appears in IEEE Transactions on Pattern Analysis and Machine Intelligence in 2013.

  7. 2017–2020

    Joining Facebook AI Research

    In September 2017, Facebook announces a new FAIR lab in Montreal, where he joins as a Research Scientist while keeping his university affiliation and devoting part of his time to supervising his graduate students. In October 2020, Mila announces that he moves from Core Academic Member to Core Industry Member after accepting a full-time research scientist position at FAIR.

  8. As of October 2026

    Research scientist at Meta FAIR

    He is a research scientist in the Fundamental AI Research (FAIR) team at Meta, an adjunct professor in the Department of Computer Science and Operations Research at the Université de Montréal, and a founding member of Mila. His current research develops theory and algorithms for representation learning that generalizes robustly out of distribution, and in 2025 he coauthors "Compositional Risk Minimization" in the Proceedings of the 42nd International Conference on Machine Learning.

Career

  1. NowMetaResearch Scientist
  2. NowUniversité de MontréalAdjunct Professor
  3. 2017–2020FAIR (Meta)Research Scientist

Education

  1. 2003Université de MontréalPhD

From the biography.

  1. 23% of people shaping AI who left a university for a company later returned to a universityIncludes Pascal Vincent. We estimate that 468 of the 2,011 people shaping AI who left a university for a company (23%) later returned to a university after a median of 5 years, and 163 of 829 (20%) made the reverse round trip.
  2. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Pascal Vincent. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Université de Montréal. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Pascal Vincent yet. Send us a free-licensed photo.

  1. Pascal Vincent - Wikidata wikidata.org. Checked 22 October 2026.
  2. A Neural Probabilistic Language Model jmlr.org. Checked 22 October 2026.
  3. Pascal Vincent | Mila mila.quebec. Checked 22 October 2026.
  4. Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion jmlr.org. Checked 22 October 2026.
  5. Pascal Vincent – CIFAR cifar.ca. Checked 22 October 2026.
  6. Representation Learning: A Review and New Perspectives arxiv.org. Checked 22 October 2026.
  7. Facebook invests over $7M U.S. in MILA and AI research in Montreal mila.quebec. Checked 22 October 2026.
  8. Mila Welcomes Five New Members to its Academic Cohort mila.quebec. Checked 22 October 2026.