Apple
Zürich, Switzerland
+32 497 06 91 46
+41 76 722 83 75

firstname.name@gmail.com

Antoine Wehenkel

Machine Learning Research Scientist

One important idea is that science is a means whereby learning is achieved, not by mere theoretical speculation on the one hand, nor by the undirected accumulation of practical facts on the other, but rather by a motivated iteration between theory and practice.       Georges EP Box

About me

I am a Machine Learning Research Scientist at Apple, within the Health AI team. I work on developing new deep learning algorithms that integrate formal domain expertise, such as that defined by scientific simulators, with real-world data to advance the design of novel sensing technologies for health. From November 2022 to November 2023, I served as a postdoctoral researcher at Apple, advised by Jörn-Henrik Jacobsen (Health AI) and Marco Cuturi (ML Research). In line with my current focus, I investigated strategies for deriving robust simulation-based inference algorithms to address issues with misspecified simulators, particularly in their application to health technologies. Prior to joining Apple, I completed my PhD in Computer Science with an FNRS Research Fellowship, under the supervision of Professor Gilles Louppe at the University of Liège, Belgium, in October 2022.

I earned my M.Sc. in Computer Engineering from the University of Liège in 2018, spending my final year as an exchange student at the École Polytechnique Fédérale de Lausanne (EPFL). At EPFL, I conducted my master's thesis in Jean-Yves Le Boudec's lab, focusing on estimating parameters of electrical distribution networks.

My vision is to enhance the interplay between fundamental sciences and machine learning techniques, both to spur scientific discovery and to develop predictive models that can be reliably deployed in the real world. In pursuit of this goal, I actively work on expanding the fields of application of simulation-based inference methods by enhancing their robustness to model misspecification and improving their integration with real-world data.

My research interests are in deep probabilistic modeling, biophysical sensor design, and simulation-based inference.

If you feel your research agenda and mine could be a good match, feel free to reach out to me!

Download CV

Publications

In Preparation

Peer Reviewed

2023
  • Robust Hybrid Learning With Expert Augmentation

    Transactions on Machine Learning Research
  • Calibrating Neural Simulation-Based Inference with Differentiable Coverage Probability

    Neural Information Processing Systems 2023
  • Distributional Reinforcement Learning with Unconstrained Monotonic Neural Networks

    Neurocomputing
  • 2022
  • Inductive Bias In Deep Probabilistic Modelling

    PhD Thesis, ULiege press
  • Towards Reliable Simulation-Based Inference with Balanced Neural Ratio Estimation

    Neural Information Processing Systems 2022
  • Averting a crisis in simulation-based inference

    Transactions on Machine Learning Research
  • A deep generative model for probabilistic energy forecasting in power systems: normalizing flows

    Applied Energy
  • 2021
    2020
    2019 and before...

    Presentations

    And posters

    Inductive Bias in Deep Probabilistic Modelling

    PhD Dissertation - University of Liège
    November 2022
    Link to the slides

    The Symbiosis between Deep Probabilistic and Scientific Models

    Presentation - HES Geneva
    October 2022
    Link to the slides

    The Symbiosis between Deep Probabilistic and Scientific Models

    Presentation - Gen U 2022, Copenhagen
    September 2022
    Link to the slides

    Normalizing Flows and Bayesian networks

    Presentation - HES Geneva
    February 2022
    Link to the slides

    Normalizing Flows and Bayesian networks

    Presentation - CogSys seminar, DTU
    October 2020
    Link to the slides Link to the video

    Normalizing Flows for Probabilistic Modeling and Inference

    Presentation - ML Journal Club, ULiège
    April 2020
    Link to the slides

    Unconstrained Monotonic Neural Networks

    Poster - NeurIPS 2019 @ Vancouver
    December 2019
    Link to the poster Link to the paper

    Neural Likelihood-Free Inference

    Presentation - Grappa, UvA in Amsterdam
    November 2019
    Link to the slides

    Unconstrained Monotonic Neural Networks

    Presentation - Benelearn 2019 @ Brussels
    November 2019
    Link to the slides

    Unconstrained Monotonic Neural Networks

    Poster - Prairie AI Summer School 2019 @ Paris
    October 2019
    Link to the poster

    Recurrent Machines For Likelihood Free Inference

    Poster - MetaLearn Workshop @ NeurIPS 2018, Montreal
    December 2018
    Link to the poster

    Readings

    When I find the time, I like reading fiction books. In particular, I often prefer books that treat about Science, Geopolitics and/or History in an accessible way.
    Over the last years my thoughts about Nature have been influenced by two books from José Rodrigues dos Santos: The Einstein Enigma: A Novel and Spinoza - L'homme qui a tué Dieu (I cannot find the english version).
    I also highly recommend these books to anyone seeking for brain entertainment, in decreasing chronological order of reading:
    • Black-out, by Marc Elsberg;
    • Antifragile: Things That Gain From Disorder, by Nassim Nicholas Taleb;
    • La Peste, by Albert Camus.