Workshop
Pitfalls of limited data and computation for Trustworthy ML
Amartya Sanyal · Alexandru Tifrea · Ankit Pensia · Franziska Boenisch · Varun Kanade · Fanny Yang · Prateek Jain · Sara Hooker · Jamie Morgenstern
MH2
Fri 5 May, midnight PDT
Machine Learning (ML) algorithms are known to suffer from various issues when it comes to their trustworthiness. This can hinder their deployment in sensitive application domains in practice. But how much of this problem is due to limitations in available data and/or limitations in compute (or memory)? In this workshop, we will look at this question from both a theoretical perspective, to understand where fundamental limitations exist, and from an applied point of view, to investigate which issues we can mitigate by scaling up our datasets and computer architectures.
Chat is not available.
Timezone: America/Los_Angeles
Schedule
Fri 12:00 a.m. - 12:10 a.m.
|
Introduction and Opening Remarks
(
Opening Remarks
)
>
SlidesLive Video |
🔗 |
Fri 12:10 a.m. - 12:45 a.m.
|
Towards neural networks robust to distribution shifts (Praneeth Netrapalli)
(
Invited Talk + Q&A
)
>
SlidesLive Video |
🔗 |
Fri 12:45 a.m. - 1:20 a.m.
|
What Neural Networks Memorize and Why (Vitaly Feldman)
(
Invited Talk + Q&A
)
>
SlidesLive Video |
🔗 |
Fri 1:25 a.m. - 1:35 a.m.
|
Beyond Confidence: Reliable Models Should Also Quantify Atypicality (Oral)
(
Oral
)
>
link
SlidesLive Video |
Mert Yuksekgonul · Linjun Zhang · James Y Zou · Carlos Guestrin 🔗 |
Fri 1:35 a.m. - 1:45 a.m.
|
On the Efficacy of Differentially Private Few-shot Image Classification (Oral)
(
Oral
)
>
link
SlidesLive Video |
Marlon Tobaben · Aliaksandra Shysheya · John Bronskill · Andrew Paverd · Shruti Tople · Santiago Zanella-Beguelin · Richard E Turner · Antti Honkela 🔗 |
Fri 1:45 a.m. - 1:55 a.m.
|
Practical Differentially Private Hyperparameter Tuning with Subsampling (Oral)
(
Oral
)
>
link
SlidesLive Video |
Antti Koskela · Tejas Kulkarni 🔗 |
Fri 1:55 a.m. - 2:05 a.m.
|
Error Discovery by Clustering Influence Embeddings (Oral)
(
Oral
)
>
link
SlidesLive Video |
Fulton Wang · Julius Adebayo · Sarah Tan · Diego Garcia-Olano · Narine Kokhlikyan 🔗 |
Fri 2:05 a.m. - 2:15 a.m.
|
Coffee Break
|
🔗 |
Fri 2:15 a.m. - 3:15 a.m.
|
Poster Session
(
Poster Session
)
>
|
🔗 |
Fri 3:15 a.m. - 4:40 a.m.
|
Lunch Break
|
🔗 |
Fri 4:40 a.m. - 5:15 a.m.
|
Impacts of Data Scarcity on Groups and Harnessing LLMs for Solution (Fereshte Khani)
(
Invited Talk + Q&A
)
>
SlidesLive Video |
🔗 |
Fri 5:15 a.m. - 5:50 a.m.
|
How (not) to Model an Adversary (Ruth Urner)
(
Invited Talk
)
>
SlidesLive Video |
🔗 |
Fri 5:50 a.m. - 6:25 a.m.
|
Practical poisoning of machine learning models (Nicholas Carlini)
(
Invited Talk
)
>
SlidesLive Video |
🔗 |
Fri 6:25 a.m. - 6:35 a.m.
|
Coffee Break
|
🔗 |
Fri 6:35 a.m. - 7:05 a.m.
|
Panel Discussion
(
Discussion Panel
)
>
SlidesLive Video |
🔗 |
Fri 7:05 a.m. - 7:15 a.m.
|
Project with Source, Probe with Target: Extracting Useful Features for Adaptation to Distribution Shifts (Oral)
(
Oral
)
>
link
SlidesLive Video |
Annie Chen · Yoonho Lee · Amrith Setlur · Sergey Levine · Chelsea Finn 🔗 |
Fri 7:15 a.m. - 7:25 a.m.
|
Efficient Utilization of Pre-Trained Model for Learning with Noisy Labels (Oral)
(
Oral
)
>
link
SlidesLive Video |
Jongwoo Ko · Sumyeong Ahn · Se-Young Yun 🔗 |
Fri 7:25 a.m. - 7:30 a.m.
|
Closing Remarks
(
Closing Remarks
)
>
|
🔗 |
Fri 7:30 a.m. - 9:00 a.m.
|
Poster Session
(
Poster Session
)
>
|
🔗 |
-
|
DORA: Exploring outlier representations in Deep Neural Networks ( Poster ) > link | Kirill Bykov · Mayukh Deb · Dennis Grinwald · Klaus R Muller · Marina Höhne 🔗 |
-
|
GeValDi: Generative Validation of Discriminative Models ( Poster ) > link | Vivek Palaniappan · Matthew Ashman · Katherine Collins · Juyeon Heo · Adrian Weller · Umang Bhatt 🔗 |
-
|
On Gradients of Deep Generative Models for Representation-Invariant Anomaly Detection ( Poster ) > link | Sam Dauncey · Christopher Holmes · Christopher Williams · Fabian Falck 🔗 |
-
|
Training, Architecture, and Prior for Deterministic Uncertainty Methods ( Poster ) > link | Bertrand Charpentier · Chenxiang Zhang · Stephan Günnemann 🔗 |
-
|
Fairness-Aware Data Valuation for Supervised Learning ( Poster ) > link | José Pombal · Pedro Saleiro · Mario Figueiredo · Pedro Bizarro 🔗 |
-
|
Learning Unforeseen Robustness from Out-of-distribution Data Using Equivariant Domain Translator ( Poster ) > link | Sicheng Zhu · Bang An · Furong Huang · Sanghyun Hong 🔗 |
-
|
ActiveLab: Active Learning with Re-Labeling by Multiple Annotators ( Poster ) > link | Hui Wen Goh · Jonas Mueller 🔗 |
-
|
KNIFE: Distilling Meta-Reasoning Knowledge with Free-Text Rationales ( Poster ) > link | Aaron Chan · Zhiyuan Zeng · Wyatt Lake · Brihi Joshi · Hanjie Chen · Xiang Ren 🔗 |
-
|
Privately Customizing Prefinetuning to Better Match User Data in Federated Learning ( Poster ) > link | Charlie Hou · Hongyuan Zhan · Akshat Shrivastava · Sid Wang · Aleksandr Livshits · Giulia Fanti · Daniel Lazar 🔗 |
-
|
Robustifying Language Models with Test-Time Adaptation ( Poster ) > link | Noah McDermott · Junfeng Yang · Chengzhi Mao 🔗 |
-
|
Pitfalls in Evaluating GNNs under Label Poisoning Attacks ( Poster ) > link | Vijay Chandra Lingam · Mohammad Sadegh Akhondzadeh · Aleksandar Bojchevski 🔗 |
-
|
Enabling Calibration In The Zero-Shot Inference of Large Vision-Language Models ( Poster ) > link | Will LeVine · Benjamin Pikus · Pranav Raja · Fernando Amat 🔗 |
-
|
Label Calibration for Semantic Segmentation Under Domain Shift ( Poster ) > link | Ondrej Bohdal · Da Li · Timothy Hospedales 🔗 |
-
|
Feature-Interpretable Real Concept Drift Detection ( Poster ) > link | Pranoy Panda · Vineeth Balasubramanian · Gaurav Sinha 🔗 |
-
|
Mark My Words: Dangers of Watermarked Images in ImageNet ( Poster ) > link | Kirill Bykov · Klaus R Muller · Marina Höhne 🔗 |
-
|
Do Models see Corruption as we see? An Item Response Theory based study in Computer Vision ( Poster ) > link | Charchit Sharma · Ayan Pahari · Deepak Vijaykeerthy · Vineeth Balasubramanian 🔗 |
-
|
Concept discovery and Dataset exploration with Singular Value Decomposition ( Poster ) > link | Mara Graziani · An-phi Nguyen · Laura O'Mahony · Henning Müller · Vincent Andrearczyk 🔗 |
-
|
Distribution Aware Active Learning via Gaussian Mixtures ( Poster ) > link | Younghyun Park · Dong-Jun Han · Jungwuk Park · Wonjeong Choi · Humaira Kousar · Jaekyun Moon 🔗 |
-
|
Understanding the class-specific effects of data augmentations ( Poster ) > link | Polina Kirichenko · Randall Balestriero · Mark Ibrahim · Shanmukha Ramakrishna Vedantam · Hamed Firooz · Andrew Wilson 🔗 |
-
|
Feature Perturbation Augmentation for Reliable Evaluation of Importance Estimators ( Poster ) > link | Lennart Brocki · Neo Christopher Chung 🔗 |
-
|
Identifying Incorrect Annotations in Multi-label Classification Data ( Poster ) > link | Aditya Thyagarajan · Elias Snorrason · Curtis Northcutt · Jonas Mueller 🔗 |
-
|
In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation ( Poster ) > link | Julian Bitterwolf · Maximilian Müller · Matthias Hein 🔗 |
-
|
A Guide for Practical Use of ADMG Causal Data Augmentation ( Poster ) > link | Audrey Poinsot · Alessandro Leite 🔗 |
-
|
Robust Neural Architecture Search by Cross-Layer Knowledge Distillation ( Poster ) > link | Utkarsh Nath · Yancheng Wang · Yingzhen Yang 🔗 |
-
|
Learning with Explanation Constraints ( Poster ) > link | Rattana Pukdee · Dylan Sam · Zico Kolter · Nina Balcan · Pradeep K Ravikumar 🔗 |
-
|
Predicting Out-of-Distribution Error with Confidence Optimal Transport ( Poster ) > link | Yuzhe Lu · Zhenlin Wang · Runtian Zhai · Soheil Kolouri · Joseph Campbell · Katia Sycara 🔗 |
-
|
Max-margin Inspired Per-sample Re-weighting for Robust Deep Learning ( Poster ) > link | Ramnath Kumar · Kushal Majmundar · Dheeraj Nagaraj · Arun Suggala 🔗 |
-
|
Superhuman Fairness ( Poster ) > link | Omid Memarrast · Trong Linh Vu · Brian Ziebart 🔗 |
-
|
A Case Study on Designing Evaluations of ML Explanations with Simulated User Studies ( Poster ) > link | Ada Martin · Valerie Chen · Sérgio Jesus · Pedro Saleiro 🔗 |
-
|
Reconstructing Training Data from Multiclass Neural Networks ( Poster ) > link | Gon Buzaglo · Niv Haim · Gilad Yehudai · Gal Vardi · michal Irani 🔗 |
-
|
Self-Consistent Chain-of-Thought Distillation ( Poster ) > link | Peifeng Wang · Zhengyang Wang · Zheng Li · Yifan Gao · Bing Yin · Xiang Ren 🔗 |
-
|
FEDERATED TRAINING OF DUAL ENCODING MODELS ON SMALL NON-IID CLIENT DATASETS ( Poster ) > link | Raviteja Vemulapalli · Warren Morningstar · Philip Mansfield · Hubert Eichner · Karan Singhal · Arash Afkanpour · Bradley Green 🔗 |
-
|
On Pitfalls of Test-Time Adaptation
(
Poster
)
>
link
SlidesLive Video |
Hao Zhao · Yuejiang Liu · Alexandre Alahi · Tao Lin 🔗 |
-
|
Conservative Prediction via Transductive Confidence Minimization ( Poster ) > link | Caroline Choi · Fahim Tajwar · Yoonho Lee · Huaxiu Yao · Ananya Kumar · Chelsea Finn 🔗 |
-
|
Differentially Private Federated Few-shot Image Classification ( Poster ) > link | Aliaksandra Shysheya · Marlon Tobaben · John Bronskill · Andrew Paverd · Shruti Tople · Santiago Zanella-Beguelin · Richard E Turner · Antti Honkela 🔗 |
-
|
Zero redundancy distributed learning with differential privacy (Oral) ( Poster ) > link | Zhiqi Bu · Justin Chiu · Ruixuan Liu · Yu-Xiang Wang · Sheng Zha · George Karypis 🔗 |
-
|
How to Make Semi-Private Learning Effective ( Poster ) > link | Francesco Pinto · Yaxi Hu · Fanny Yang · Amartya Sanyal 🔗 |
-
|
Sentence Embedding Encoders are Easy to Steal but Hard to Defend ( Poster ) > link | Adam Dziedzic · Franziska Boenisch 🔗 |
-
|
Project with Source, Probe with Target: Extracting Useful Features for Adaptation to Distribution Shifts ( Poster ) > link | Annie Chen · Yoonho Lee · Amrith Setlur · Sergey Levine · Chelsea Finn 🔗 |
-
|
Efficient Utilization of Pre-Trained Model for Learning with Noisy Labels ( Poster ) > link | Jongwoo Ko · Sumyeong Ahn · Se-Young Yun 🔗 |
-
|
Beyond Confidence: Reliable Models Should Also Quantify Atypicality ( Poster ) > link | Mert Yuksekgonul · Linjun Zhang · James Y Zou · Carlos Guestrin 🔗 |
-
|
On the Efficacy of Differentially Private Few-shot Image Classification ( Poster ) > link | Marlon Tobaben · Aliaksandra Shysheya · John Bronskill · Shruti Tople · Santiago Zanella-Beguelin · Richard E Turner · Antti Honkela 🔗 |
-
|
Practical Differentially Private Hyperparameter Tuning with Subsampling ( Poster ) > link | Antti Koskela · Tejas Kulkarni 🔗 |
-
|
Error Discovery by Clustering Influence Embeddings ( Poster ) > link | Fulton Wang · Julius Adebayo · Sarah Tan · Diego Garcia-Olano · Narine Kokhlikyan 🔗 |