AI Paper Atlas

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

On Focal Loss for Class-Posterior Probability Estimation: A Theoretical Perspective

Nontawat Charoenphakdee, Jayakorn Vongkulbhisal, Nuttapong Chairatanakul, Masashi Sugiyama

CVPR 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 22 citations (OpenAlex)

Check 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 focal loss with any γ ≥ 0 instead of an uncalibrated surrogateMulticlass classification risk minimization · Multiclass classificationBayes classification riskImproved
Use focal loss with γ > 0 without transforming its outputFocal-loss classifier · Class-posterior probability estimationProbability estimation correctnessHurt
Apply the proposed Ψ^γ transformation to focal-loss scoresFocal-loss classifier · Synthetic 3-class classificationKLDImproved
Apply the proposed Ψ^γ transformation to focal-loss scoresFocal-loss classifier · Synthetic 3-class classificationClassification errorNo clear effect
Apply Ψ^γ after the softmax layerResNet models trained on CIFAR10 and SVHN · CIFAR10 and SVHNECEMixed
Apply Ψ^γ after the softmax layerResNet110 trained with label smoothing or focal-loss temperature scaling · CIFAR10 and SVHNECEImproved
Use focal loss without Ψ^γ rather than cross-entropyNeural-network classifiers · 30 datasetsECEHurt
Use focal loss without Ψ^γ rather than cross-entropyNeural-network classifiers · Some benchmark settingsECEMixed
Increase γ for focal-loss training in binary classificationFocal-loss classifier · Binary classificationMaximum-score underconfidenceHurt
Use temperature scaling rather than Ψ^γ to post-process focal-loss scoresFocal-loss classifier · Synthetic 3-class classificationClass-posterior estimationHurt
Increase focal-loss γ under label smoothing or focal-loss temperature scalingResNet110 · CIFAR10 and SVHNECEHurt
Use Ψ^γ when the focal risk minimizer is estimated less accurately, as with less training data or a more complex modelFocal-loss classifier · CIFAR10, CIFAR10-s, and SVHNECEHurt

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