AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Cost-Sensitive Self-Training for Optimizing Non-Decomposable Metrics

Harsh Rangwani, Shrinivas Ramasubramanian, Sho Takemori, Kato Takashi, Yuhei Umeda, Venkatesh Babu Radhakrishnan

NeurIPS 2022 · Label-efficient and robust learning › Semi-supervised learning · approach: Self-training theory · 1 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
Add CSST's cost-sensitive consistency training to FixMatchFixMatch · CIFAR10-LT, ρ=100Avg. Rec: 0.78 -> 0.80Improved
Add CSST's cost-sensitive consistency training to FixMatchFixMatch · CIFAR100-LT, ρ=10Min. HT Cov: 0.004 -> 0.010Improved
Add CSST's cost-sensitive consistency training to FixMatchFixMatch · ImageNet100-LT, ρ=10Min. HT Cov: 0.007 -> 0.010Improved
Add CSST's cost-sensitive consistency training to UDAUDA · IMDb, ρ=10Min Rec: 0.66 -> 0.88Improved
Add CSST's cost-sensitive consistency training to UDAUDA · IMDb, ρ=100Min Rec: 0.00 -> 0.75Improved
Add CSST's cost-sensitive consistency training to UDAUDA · DBpedia-14, ρ=100Min Rec: 0.65 -> 0.97Improved
Remove KL-based thresholding (ablation)CSST(FixMatch) · CIFAR10-LT, ρ=100Avg. Rec: 0.80 -> 0.67Hurt
Remove KL-based thresholding (ablation)CSST(FixMatch) · CIFAR100-LT, ρ=10Avg. Rec: 0.63 -> 0.47Hurt
Remove KL-based thresholding (ablation)CSST(FixMatch) · ImageNet100-LT, ρ=10Avg. Rec: 0.58 -> 0.26Hurt
Remove KL-based thresholding (ablation)CSST(FixMatch) · CIFAR10-LT, ρ=100Min. Cov: 0.092 -> 0.093Mixed
Increase the KL-divergence threshold too far (ablation)CSST(FixMatch) · CIFAR10-LT, ρ=100Avg. RecHurt
Increase the amount of labeled data while holding unlabeled data constant (ablation)CSST · Experiments in Figure [ref]Avg. RecImproved

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