AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Beyond probability partitions: Calibrating neural networks with semantic aware grouping

Jia-Qi Yang, De-Chuan Zhan, Le Gan

NeurIPS 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Probabilistic calibration · 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
Use learned semantic-aware groups with group-specific temperature scaling instead of global temperature scalingGC+TS · CIFAR10-Resnet152 constructed binary settingECE: 0.0157 -> 0.0118Improved
Use learned semantic-aware groups with group-specific calibration instead of global calibrationGC+ETS · CIFAR10, CIFAR100, Imagenettop-label ECE: -8.92%Improved
Use learned semantic-aware groups with group-specific temperature scaling instead of global temperature scalingGC+TS · CIFAR10, CIFAR100, Imagenettop-label ECE: -5.03%Improved
Use learned group partitions and group-specific calibrationGC+TS · CIFAR10-Resnet152 constructed binary settingPCE: 0.0174 -> 0.0141Improved
Increase the number of partitions (ablation)GC+TS · CIFAR10-Resnet152calibration performanceImproved
Decrease the number of groups within each partition (ablation)GC+TS · CIFAR10-Resnet152NLL and ECEHurt
Set grouping-function regularization too high (value of 1) (ablation)GC+TS · CIFAR10-Resnet152calibration performance: 1Hurt
Set grouping-function regularization too low (ablation)GC+TS · CIFAR10-Resnet152calibration performanceHurt
Calibrate using the validation set rather than the hold-out calibration setexisting calibration methods · calibration experimentscalibration performanceHurt
Train the grouping function on the validation set and transfer it to the hold-out setGC+TS · calibration experimentscalibration performanceImproved

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