AI Paper Atlas

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

PiCO+: Contrastive Label Disambiguation for Robust Partial Label Learning

Haobo Wang, Ruixuan Xiao, Yixuan Li, Lei Feng, Gang Niu, Gang Chen, Junbo Zhao

TPAMI 2023 · Label-efficient and robust learning › Semi-supervised learning · approach: Confidence-based pseudo-label selection · 90 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 contrastive loss (with uniform pseudo targets) (ablation)PiCO · CIFAR-10 (q=0.5)accuracy: 76.46 -> 84.50Improved
add prototype-based label disambiguation (ablation)PiCO · CIFAR-10 (q=0.5)accuracy: 84.50 -> 93.58Improved
add prototype-based label disambiguation (ablation)PiCO · CIFAR-100 (q=0.05)accuracy: 64.11 -> 72.74Improved
use hard prototype targets with moving average instead of soft prototype probabilities (ablation)PiCO · CIFAR-10 (q=0.5)accuracy: 81.67 -> 93.58Improved
add moving-average smoothing (phi>0) to one-hot prototype assignment (ablation)PiCO · CIFAR-100 (q=0.05)accuracy: 70.10 -> 72.74Improved
use distance-based clean selection instead of small-loss selection (ablation)PiCO+ · Noisy PLL CIFAR-10 (q=0.5, eta=0.2)accuracy: 91.22 -> 92.59Improved
add semi-supervised training on noisy samples beyond clean-only PiCO (ablation)PiCO+ · Noisy PLL CIFAR-10 (q=0.5, eta=0.2)accuracy: 87.63 -> 92.59Improved
add prototype-based label guessing loss L_n-cls (ablation)PiCO+ · Noisy PLL CIFAR-10 (q=0.5, eta=0.2)accuracy: 91.14 -> 92.59Improved
add label-driven noisy contrastive loss L_n-cont (ablation)PiCO+ · Noisy PLL CIFAR-10 (q=0.5, eta=0.2)accuracy: 90.87 -> 92.59Improved
add kNN neighbor-augmented contrastive loss (ablation)PiCO+ · Noisy PLL CIFAR-10 (q=0.5, eta=0.2)accuracy: 87.43 -> 92.59Improved
add mixup training (ablation)PiCO+ · Noisy PLL CIFAR-100 (q=0.05, eta=0.2)accuracy: 68.51 -> 72.98Improved
raise prototype momentum gamma from 0.99 to 0.999 (ablation)PiCO · CIFAR-100 (q=0.05)accuracy: 72.74 -> 67.33Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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