AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Online Label Shift: Optimal Dynamic Regret meets Practical Algorithms

Dheeraj Baby, Saurabh Garg, Tzu-Ching Yen, Sivaraman Balakrishnan, Zachary Chase Lipton, Yu-Xiang Wang

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use FLH-FTL label marginal estimation instead of FTHonline label shift correction · CIFAR UOLS Bernoulliclassification error (%): 11 -> 10Improved
use FLH-FTL instead of FTHonline label shift correction · Synthetic UOLS Bernoulliclassification error (%): 6.5 -> 5.4Improved
use FLH-FTL instead of FTHonline label shift correction · Fashion UOLS Bernoulliclassification error (%): 8.5 -> 7.7Improved
use FLH-FTL instead of FTHonline label shift correction · MNIST and EuroSAT/ArXiv Sinusoidal shiftclassification error (%): 3.3 -> 3.3No clear effect
use FLH-FTL instead of FTHonline label shift correction · CIFAR UOLS BernoulliMSE of label marginal: 0.27 -> 0.17Improved
use FLH-FTL instead of ROGDonline label shift correction · CIFAR UOLS BernoulliMSE of label marginal: 0.39 -> 0.17Improved
decrease amount of offline holdout data (ablation)FLH-FTL vs alternatives · CIFAR10 UOLS monotone shiftclassification error and marginal MSENo clear effect
increase offline training data at expense of holdout (ablation)FLH-FTL · CIFAR10 UOLS monotone shiftclassification errorImproved
continual training with re-sampling (CT-RS) instead of naive continual trainingSOLS baseline · CIFAR10/MNIST SOLS Bernoulliclassification performanceImproved
CT-RS instead of weighted ERM from scratchSOLS · MNIST/CIFAR SOLScomputation cost: 5--15x fasterImproved
FLH-FTL instead of FTH for marginal estimation in SOLSCT-RS · MNIST/CIFAR SOLSmarginal estimation and classificationImproved
use FLH-FTL instead of ATLAS/UOGDonline label shift correction · CIFAR UOLS Bernoulliclassification error (%): 13 -> 10Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy