Morgane Goibert

PhD student in ML/Deep Learning @CriteoAILab and @IP_Paris_ / @TelecomParis, specialized in Adversarial Robustness problems and related stuff.

[David Stutz @davidstutz92]

To kick off my blog article series on adversarial examples and adversarial training, I just published the first three articles that lay some of the foundations: monitoring training and good clean performance - 2.56% test error on CIFAR10. Start here: https://t.co/S7OXjxEvD9

[Morgane Goibert @MGoibert]

Check out Elvis's presentation of our paper at @aistats_conf yesterday 👇

[Elvis Dohmatob @dohmatobelvis]

Here are the slides for "Origins of Low-Dimensional Adversarial Perturbations" https://t.co/qJTdXXkTpW

[Criteo AI Lab @CriteoAILab]

Do you want to know how to generate fluent text from structured data and vice versa by leveraging heterogeneous data sources?👀
Don’t miss the opportunity to discuss it with @songdng today at @aistats_conf🙌
⏰Auditorium 1 Foyer 6 - 4:30pm - 7pm
https://t.co/0bl75NVyO1
#aistats23

[Imad Aouali @imad_aouali]

🥳🎉 Our paper "Exponential Smoothing for Off-Policy Learning" has been accepted at #ICML2023
@icmlconf for an *Oral Presentation*.

Many thanks to my co-authors David Rohde (@CriteoAILab), Victor-Emmanuel Brunel and @Korba_Anna (@Ensaeparis/@CrestUmr). 🤩🤩 See you in Hawaii!

[Morgane Goibert @MGoibert]

Today at @aistats_conf, @dohmatobelvis will present our paper "Origins of Low-Dimensional Adversarial Perturbations" in Oral session 2😃
Check it here 👉 https://t.co/mSfVL9jaWc

The paper studies the conditions under which low-dimensional black box attacks succeed 😎

[ICML Conference @icmlconf]

#ICML2023 Workshops have been posted: https://t.co/wDVzYs7oAj

[Morgane Goibert @MGoibert]

@rballeba This is definitely a good idea. IMHO proposing a book would help the field structure itself and gain more traction 😀

[Morgane Goibert @MGoibert]

@qberthet @cjdamato As an AC, do you have statistics about that (eg ratio of replies before any intervention from you, etc.) over a sufficient nb of reviewers? 🧐 This would be very interesting to conduct such a study to get a broader view. On my paper, I got just one reply over 3 reviews.

[Morgane Goibert @MGoibert]

@PierreAlquier Ahahah, good one 😅 (though I don't know what's the more depressing, seeing a repetition of the same sentence or all the variants!)

[Morgane Goibert @MGoibert]

If the reviewers for my paper that asked for clarifications/experiments that we indeed provided in the rebuttal (after working quite a lot during a week-end, remember this very good timeline from @icmlconf ) could see this...😅

[Murat Kocaoglu @murat_kocaoglu_]

If you are reviewing for ICML, please pay attention to the rebuttals of papers you gave low scores to and engage with the authors.

It takes 5 min of your time to correct a misunderstanding that will save someone else's 1-year worth of work.

#ICML2023

[Murat Kocaoglu @murat_kocaoglu_]

If you are reviewing for ICML, please pay attention to the rebuttals of papers you gave low scores to and engage with the authors.

It takes 5 min of your time to correct a misunderstanding that will save someone else's 1-year worth of work.

#ICML2023

[Morgane Goibert @MGoibert]

@Pseudomanifold @LogConference My experience with reviewers has never been great, but clearly, #ICML2023 has been really horrible😤 I don't know to fix it (bad/good incentives to reviewers?), anyway as a young researcher, the only incentive I get is to avoid being a reviewer as much as possible. Sad🫤

[Morgane Goibert @MGoibert]

🥇Results: much better accuracy on underspecified problems 🔥even in cases where labels are completely correlated with spurious features🔥

👉 Such an OOD setting is quite similar to the adversarial one. I wonder if such strategies can be adapted to it 🧐

[Morgane Goibert @MGoibert]

Morning read [1/3]

💡Training multiple NNs heads while enforcing diversity of predictions ensure great accuracy results for out-of-distribution or underspecified problems.

👉https://t.co/JUW6mu2uT3 from @yoonholeee @HuaxiuYaoML @chelseabfinn at #iclr2023

[Morgane Goibert @MGoibert]

Main strength 🔥 the "Diversify" stage 👉 heads are forced to be different because of a term minimizing mutual information between predictions from different heads in the training loss, simple and effective👌

The best head is then selected in the "Disambiguate" stage😊

[Morgane Goibert @MGoibert]

Very nice way to accompany your paper. Great job! 👌

[Randall Balestriero @randall_balestr]

Awesome website summarizing our latest TMLR paper demonstrating how deep networks pruning can be easily explained/visualized and improved simply by formulating it in terms of the DN's spline partition!
Paper: https://t.co/30CXYmAWKM
Code: https://t.co/heGXelObVh

[Morgane Goibert @MGoibert]

@Pseudomanifold @shortstein I also include a well-written comment only when it was particularly easy to follow (especially when I don't really know some concepts used in the paper😅). It's so much easier as a reviewer compared to papers that don't make any efforts on that side 😑

[Morgane Goibert @MGoibert]

@kinisoftware @CriteoEng @NotionHQ So cool 😃 I didn't know you could use it that way, thanks for showcasing it! It can definitely save a lot of time 😎

[Morgane Goibert @MGoibert]

[3/3] The exps are done on NNs with large filters as it is harder to see something on smaller ones.
I'm curious if some structure can still be extracted (but how?) from smaller filters🧐

Anyway, very interesting to see that structure is once again important for #deeplearning 😊

[Morgane Goibert @MGoibert]

[Morning read 1/3]

💡Better accounting for the structure of NNs conv filters NNs improves performance 👉 https://t.co/7fHmb6f6IG by @ashertrockman & al. @iclr_conf

They found that conv filters are highly structured, and designed a new initialization method to improve NNs perf

[Morgane Goibert @MGoibert]

[2/3] The strength of their initialization is that it's a simple Gaussian one where just the covariance matrix is carefully crafted, meaning it's very easy to implement 🤩

Results 🧐
1⃣Improved accuracy and 2⃣faster convergence 🔥