Ching Lam CHOI

               


I'm a first year MIT EECS PhD student at CSAIL, co-supervised by Phillip Isola, Antonio Torralba and Stefanie Jegelka.

During my CUHK bachelor's, I researched with wonderful people — Aaron Courville and Yann Dauphin at MILA; Wieland Brendel at Tübingen's Max Planck Institute; Serge Belongie at Copenhagen's Pioneer Centre; CUHK's Hongsheng Li (MMLab), Anthony Man-Cho So, Farzan Farnia and Qi Dou.


CV  /  Bio [WIP]  /  Coffee-hours  /  Blog [WIP]

profile photo

Research

I am interested in probing the robustness, causal reasoning and epistemic capabilities of ML systems. In robustness, I investigate making systems run on synthetic data, such that they scale in a sustainable, privacy-preserving manner. Through causality, I aim to make models rational, understandable and reliable, to provide safety and logic-symbolic guarantees. Epistemology concerns understanding the extent of AI's capabilities, testing its limits in expression and generation through computational/complexity theory.

UNI: Unlearning-based Neural Interpretations UNI: Unlearning-based Neural Interpretations
Ching Lam Choi, Alexandre Duplessis, Serge Belongie
Preprint, 2024   [cite]

Debiased path attribution by perturbing in the unlearning direction of steepest ascent.
On the Generalization of Gradient-based Neural Network Interpretations On the Generalization of Gradient-based Neural Network Interpretations
Ching Lam Choi, Farzan Farnia
Preprint, 2023   [cite]

Gradient-based XAI fails to generalise from train to test data: regularisation crucially helps.
Universal Adversarial Directions Universal Adversarial Directions
Ching Lam Choi, Farzan Farnia
Preprint, 2022   [cite]

UAD is an adversarial examples game with a pure Nash equilibrium and better transferability.
Self-distillation with Batch Knowledge Ensembling Improves ImageNet Classification Self-distillation with Batch Knowledge Ensembling Improves ImageNet Classification
Yixiao Ge, Xiao Zhang, Ching Lam Choi, Ka Chun Cheung, Peipei Zhao, Feng Zhu, Xiaogang Wang, Rui Zhao, Hongsheng Li
Preprint, 2021   [cite]

BAKE distils self-knowledge of a network, by propagating samples similarities in each batch.
DivCo: Diverse Conditional Image Synthesis via Contrastive Generative Adversarial Network DivCo: Diverse Conditional Image Synthesis via Contrastive Generative Adversarial Network
Rui Liu, Yixiao Ge, Ching Lam Choi, Hongsheng Li
CVPR, 2021   [cite]

DivCo uses contrastive learning to reduce mode collapse in GANs.

News, Talks, Events
[Presentation 18.11.2024]   "Query Complexity for ML" at Jegelka lab meeting, hosted by Stefanie Jegelka.

[Presentation 06.11.2024]   "Synthetic Data for Property Testing" at Isolalab lab meeting, hosted by Phillip Isola.

[Talk 27.05.2024]   "(Un)learning to Explain" at Krueger AI Safety Lab, hosted by David Krueger.

[Talk 22.03.2024]   "Robust Scaling: Trustworthy Data" at CleverHans Lab, hosted by Nicolas Papernot.

[Presentation 08.05.2023]   "Robustness Transfer in Distillation" at Brendel & Bethge joint lab meeting, hosted by Wieland Brendel.

[Talk 16.02.2022]   "Federated learning and the Lottery Ticket Hypothesis" at ML Collective lab meeting, hosted by Rosanne Liu.

[Talk 31.07.2020]   "Julia Track Google Code In and Beyond" at JuliaCon 2020.

[Talk 24.07.2020]   "Corona-Net: Fighting COVID-19 With Computer Vision" at EuroPython 2020.

[Talk 13.06.2020]   "Julia – Looks like Python, feels like Lisp, runs like C/Fortran" at Hong Kong Open Source Conference 2020.
[Co-organiser]   Causality and Large Models Workshop @NeurIPS 2024

[Co-organiser]   New in ML Workshop @NeurIPS 2024

[Co-organiser]   New in ML Workshop @NeurIPS 2023

[Co-organiser]   CoSubmitting Summer Workshop @ICLR 2022

[Co-organiser]   Undergraduates in Computer Vision Social @ICCV 2021

[Reviewer]   CVPR 2023, 2024; ICCV 2023; NeurIPS 2023, 2024; ICLR 2024, 2025; ICML 2024, 2025; ECCV 2024*; ACCV 2024; AAAI 2025; AISTATS 2025 (* indicates outstanding reviewer award).

Hobbies

Stephen Fry: "We are not nouns, we are verbs. I am not a thing [but] a person who does things."

I am an amateur of many trades for my own amusement. I’m easily moved by words, music, nature, Hugh Laurie; utterly immovable by (non-climbing) gym memberships, organised religion and oatmeal. Besides “A Bit of Fry and Laurie”, I enjoy:

Tübingen, the Neckar and the Hölderlinturm Sports and Nature

I boulder, climb, hike, fence, play volleyball and would open a climbing gym called “Sloblock”.
Picture with Hilary Hahn post Mendelssohn's Violin Concerto in E minor in Dortmund Music

I adore the cello, oboe, French horn, voice and violin. Favourite artists include the Berliner Philharmoniker, the 12 Cellists of the Berlin Phil, Tom Rosenthal, Joni Mitchell, Sibelius, Samuel Barber, Augustin Hadelich, Hilary Hahn, Bruno Delepelaire, Ludwig Quandt, Nathan Chan, Yo-Yo Ma, Cornelius Meister, Jonathan Kelly.
Picture taken at the Deutsches Romantik-Museum Literature

I travel with comfort books and particularly adore Stefan Zweig ("Chess", "The World of Yesterday"), Alice Munro ("Friend of My Youth"), Franz Kafka ("The Castle"), Oscar Wilde ("Happy Prince", "Lady Windermere's Fan", "A Woman of No Importance"), F. Scott Fitzgerald ("Gatsby") and Shakespeare ("Twelfth Night", "Taming of the Shrew", "Richard III", "Hamlet", "Macbeth", "Othello"); "A Time of Gifts" by Patrick Leigh Fermor.

Website design credits to Jon Barron.