Ferry Julien

[Thibaut Vidal @vidalthi]

Happy to announce that our paper was rejected as a spotlight (5/5/4) at #ICML2026.

If the methodology was complex enough to confuse the metareviewer, perhaps it may still be of broader interest to you 🙂.

Happy to discuss the work if you are into optimal counterfactual maps that permit explanations in milliseconds, or into the occasional ups and downs of academic publishing 🚣

[Thibaut Vidal @vidalthi]

Hot off the press! Our research on the *risks of training-data reconstruction from random forest models* has just been nominated as one of @QuebecScience's Top 10 Scientific Discoveries of 2024! 🌟 🚀
https://t.co/GFwiw7am32
https://t.co/VgAjkBtsNX

[Schwartz Reisman Institute @TorontoSRI]

Want to know what's happening at @satml_conf today? Luckily we have @thegautamkamath of @UWaterloo on the ground reporting for us!

First up: this morning's session on privacy chaired by @fraboeni feat. work by @FerryJulien12,@hjy836, and others. A thread... 🧵#SaTML2024 (1/5)

[SaTML Conference @satml_conf]

We're onto the technical sessions. First up, privacy! @FerryJulien12 presents about data reconstruction risks of interpretable models.

[Thibaut Vidal @vidalthi]

*Trained Random Forests Completely Reveal your Dataset!* https://t.co/hMmDkiNFSe
#MachineLearning models encode extensive information that can be exploited to reconstruct most --if not all-- of their private training data. Wanna know how? Buckle up! (1/8)

[Thibaut Vidal @vidalthi]

Join us today at 3pm EST for a @DS4DM & SCALE-AI Chair seminar from Maximilian Schiffer about "Combinatorial optimization augmented machine learning for contextual multi-stage problems". Participation is also possible via Zoom! https://t.co/EC2o8aacD1
#ORMS #DecisionAwareLearning

[Ferry Julien @FerryJulien12]

Curious about the details ? Check out our preprint on arXiv (https://t.co/x8vMUxTgxA)!
This is joint work with Gabriel Laberge and @umaivodj

[Ferry Julien @FerryJulien12]

You want to learn an interpretable ML model, but you are afraid of the potential performance harm compared to a more complex, black-box model ?

Why not learn a Hybrid Interpretable Model ?

[Ferry Julien @FerryJulien12]

To this end, you can use our new Python module, named HybridCORELS and available on PyPI (https://t.co/bKbFUAvw12) and GitHub (https://t.co/FZk2n0FKrX).

[Ferry Julien @FerryJulien12]

Our extensive empirical study demonstrates that this method effectively improves fairness generalization (on unseen data).

[Ferry Julien @FerryJulien12]

Happy to share with you our work presented last month at the #aaai23 conference (Journal Track, and CP-ML Bridge) in Washington.
Paper: https://t.co/quWY1LFgGD
Recording of the presentation, slides and poster: https://t.co/nxcZs6kp3y

[Ferry Julien @FerryJulien12]

We generalize this intuition and propose to learn models that are fair on a given dataset, but also on all its subsets closer than a defined Jaccard distance.
To measure or enforce such robustness notion, we formulate an Integer Programming model.

[Ferry Julien @FerryJulien12]

Just presented our work at @satml_conf in Raleigh (NC, USA) ! This is joint work with @umaivodj, @siala__, Sébastien Gambs and Marie-José Huguet

[Ferry Julien @FerryJulien12]

Great news! Our @MLJ_Social paper "Improving Fairness Generalization Through a Sample-Robust Optimization Method" (https://t.co/quWY1LFgGD) is accepted for presentation in the #AAAI23 Journal Track.
Joint work with @umaivodj, Sébastien Gambs, Marie-José Huguet, and @siala__.

[Ferry Julien @FerryJulien12]

And source code on GitHub: https://t.co/oageWvdVoS

[Ferry Julien @FerryJulien12]

Learning fair models can endanger the privacy of sensitive attributes.
This is illustrated in our paper "Exploiting Fairness to Enhance Sensitive Attributes Reconstruction", accepted to @satml_conf.
Joint work with @umaivodj, Sébastien Gambs, Marie-José Huguet, and @siala__.

[Ferry Julien @FerryJulien12]

@satml_conf @umaivodj @siala__ Preprint is available on arXiv: https://t.co/tn1WdQ2oXz