Thang Doan

Pr. Applied Scientist @Oracle | Postdoc @Mila_Quebec, PhD @mcgillu, MSc @ucl

[Thang Doan @doan_tl]

Delighted to share our work “A Theoretical Analysis of Catastrophic Forgetting through the NTK Overlap Matrix “ accepted at AISTATS 2021 w/ @Mehdi_A_Bennani, @bogdan_mazoure, @grwip, @PierreAlquier

Paper: https://t.co/4eGWg5JXfN
Code: https://t.co/wD70HUIZ00

(1/n) 👇

[Jean-Baptiste Gaya @jb_gaya]

I will present our paper Building a Subspace of Policies for Scalable Continual Learning at #ICLR23 at 3:10pm in room AD10. If you're interested in the topic, come to meet and chat about it !
@doan_tl @LucasPCaccia @LudovicDenoyer @LaureSoulier @robertarail

[Roberta Raileanu @robertarail]

CSP accepted as spotlight at #ICLR2023. Kudos to @jb_gaya who led this project and to all the other wonderful co-authors @doan_tl @LucasPCaccia @LaureSoulier @LudovicDenoyer

[Roberta Raileanu @robertarail]

Building a Subspace of Policies for Continual Reinforcement Learning

When: Fri Dec 9, 8:25am PST
Where: Virtual

Paper: https://t.co/g5A6RMV0jN
Website: https://t.co/TNdejzQ0Rx
Thread: https://t.co/HdinKu36zi

led @jb_gaya w @doan_tl @LucasPCaccia @LaureSoulier @LudovicDenoyer

[Thang Doan @doan_tl]

[Worskhop on Theory of Continual learning (ICML'21)]

Thrilled to announce our awesome speaker line up: https://t.co/Rprjm7MFLY

You can ask questions to our speakers here:
https://t.co/Y2g4DuveRC

Happening this Friday 23rd !🔥🚀

[Thang Doan @doan_tl]

We [EXCEPTIONALLY] extend the deadline of our workshop on "Theory of Continual Learning" (ICML'21) to June 11th (AoE)
more information at:
https://t.co/Rprjm7v4no

[Thang Doan @doan_tl]

We have extended the deadline of our workshop on "Theory and Foundation of Continual Learning", ICML'21
to June 07th (AoE)

website: https://t.co/Rprjm7MFLY

[Thang Doan @doan_tl]

If you want to know more about Self-supervision methods in RL, join our workshop today !
Happening in 1 hour 🔥🎉

[Ankesh Anand @ankesh_anand]

Quick reminder that the deadline for the Self-Supervision for RL (https://t.co/LGNnmFAsMt) workshop at ICLR'21 is tomorrow (Feb 26th) AoE!

Looking forward to everyone's submissions, and feel free to email us with any questions!

[Thang Doan @doan_tl]

Glad to share my talk at the @ContinualAI reading group hosted by @v_lomonaco ! :)

[Vincenzo Lomonaco @v_lomonaco]

🔥 The recording of the latest @ContinualAI seminar with @doan_tl is here! 🔥

https://t.co/hkzWYuLv7O

[Thang Doan @doan_tl]

@LucasPCaccia thanks Lucas ! 😀

[Thang Doan @doan_tl]

Delighted to share our work “A Theoretical Analysis of Catastrophic Forgetting through the NTK Overlap Matrix “ accepted at AISTATS 2021 w/ @Mehdi_A_Bennani, @bogdan_mazoure, @grwip, @PierreAlquier

Paper: https://t.co/4eGWg5JXfN
Code: https://t.co/wD70HUIZ00

(1/n) 👇

[Thang Doan @doan_tl]

We propose a variant of Orthogonal Gradient Descent (OGD) which leverages the structure of the data through Principal Component Analysis called PCA-OGD. Our method is twice as memory efficient as OGD.

(4/n)

[Thang Doan @doan_tl]

We introduce the “NTK Overlap matrix” as a task similarity measure that governs the CF. Unlike Vanilla SGD, projection based methods mitigate CF by operating orthogonally to previous task subspaces. This reduces the overlap between tasks leading to less forgetting.

(3/n) 👇

[CIFAR @CIFAR_News]

Canada welcomes 29 newly-appointed Canada CIFAR AI Chairs at @AmiiThinks @Mila_Quebec @VectorInst - congratulations to all! #AICan
https://t.co/dBhAj5PtVp

[Ankesh Anand @ankesh_anand]

The RL formalism is powerful in its generality, but poses a hard problem: how can we design agents that learn efficiently & generalize well, given only sensory info and a reward signal?

Self-supervision might be the answer, join us at the ICLR workshop: https://t.co/lj7Q30n1qr

[Sébastien Couasnon  @SebCouasnon]

Population déjà vaccinée :
🇮🇱24%
🇬🇧5,5%
🇺🇸4%
🇩🇰2,5%
🇮🇹1,7%
🇪🇸1,6%
🇨🇦1,3%
🇩🇪1,2%
🇪🇺1,1%
🇵🇹1%
🇷🇺1%
🇫🇷0,6%

[Mehdi Bennani @MehdiBennaniML]

Thanks for hosting the reading group and for the discussions @v_lomonaco :)

[Vincenzo Lomonaco @v_lomonaco]

🔥 The @ContinualAI Reading Groups are back! 🔥

This Friday, 17.30 CET we will host Mehdi Abbana Bennani (Aqemia) who will present the paper: "Generalisation Guarantees for Continual Learning with Orthogonal Gradient Descent"

https://t.co/xySe0zcCTc