AI Paper Atlas

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

Rethinking Calibration of Deep Neural Networks: Do Not Be Afraid of Overconfidence

Deng-Bao Wang, Lei Feng, Min-Ling Zhang

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Confidence calibration objectives · 39 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 regularization (label smoothing, L1 norm, focal loss) in training instead of plain cross-entropyResNet-32 · CIFAR-10ECE without post-hoc calibration: 6.43 -> 2.72 (label smoothing)Improved
add regularization (label smoothing) in training then apply temperature scalingResNet-32 · CIFAR-10ECE with TS: 0.95 -> 2.54Hurt
add focal loss in training then apply temperature scalingResNet-32 · CIFAR-100ECE with TS: 1.35 -> 2.14Hurt
add L1 norm regularization then apply temperature scaling8-layer 1D-CNN · 20 NewsgroupsECE with TS: 3.11 -> 2.71Improved
add label smoothing or focal loss then apply histogram binningResNet-32 · CIFAR-10ECE with HB: 0.74 -> 0.94 (LS); 0.74 -> 1.65 (FL)Hurt
add focal loss with strong coefficientResNet-32 · CIFAR-10Accuracy: 90.46 -> 87.84Hurt
replace cross-entropy with inverse focal loss (weights confident samples more) plus temperature scalingResNet-32 · CIFAR-10 and CIFAR-100ECE after TSImproved
replace cross-entropy with inverse focal loss, histogram binningResNet-32 · SVHN, CIFAR-10, CIFAR-100ECE after HBMixed
replace cross-entropy with inverse focal loss, no post-hoc calibrationResNet-32 · most datasetsECE without post-hocHurt
replace cross-entropy with inverse focal loss, large coefficient8-layer 1D-CNN · 20 NewsgroupsAccuracyHurt

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