Lune

NeurIPS2021Top-tier venue

A Geometric Perspective towards Neural Calibration via Sensitivity Decomposition

Junjiao Tian, Dylan Yung, Yen-Chang Hsu, Zsolt Kira

Abstract

It is well known that vision classification models suffer from poor calibration in the face of data distribution shifts. In this paper, we take a geometric approach to this problem. We propose Geometric Sensitivity Decomposition (GSD) which decomposes the norm of a sample feature embedding and the angular similarity to a target classifier into an instance-dependent and an instance-independent component. The instance-dependent component captures the sensitive information about changes in the input while the instance-independent component represents the insensitive information serving solely to minimize the loss on the training dataset. Inspired by the decomposition, we analytically derive a simple extension to current softmax-linear models, which learns to disentangle the two components during training. On several common vision models, the disentangled model outperforms other calibration methods on standard calibration metrics in the face of out-of-distribution (OOD) data and corruption with significantly less complexity. Specifically, we surpass the current state of the art by 30.8% relative improvement on corrupted CIFAR100 in Expected Calibration Error. Code available at https: //github.com/GT-RIPL/Geometric-Sensitivity-Decomposition.git .

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

4 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Adaptive Calibrator Ensemble: Navigating Test Set Difficulty in Out-of-Distribution Scenarios cites Uncertainty Estimation Using a Single Deep Deterministic Neural Network
  • Adaptive Calibrator Ensemble: Navigating Test Set Difficulty in Out-of-Distribution Scenarios cites Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness
  • Calibrating Multimodal Learning cites Uncertainty Estimation Using a Single Deep Deterministic Neural Network
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers4

  1. Better Uncertainty Calibration via Proper Scores for Classification and BeyondSebastian G. Gruber, Florian BuettnerNeurIPS 2022 · 88 citations
  2. Calibrating Multimodal LearningHuan Ma, Qingyang Zhang, Changqing Zhang, Bingzhe Wu et al.ICML 2023 · 42 citations
  3. Adaptive Calibrator Ensemble: Navigating Test Set Difficulty in Out-of-Distribution ScenariosYuli Zou, Weijian Deng, Liang ZhengICCV 2023 · 12 citations
  4. Trainable Projected Gradient Method for Robust Fine-TuningJunjiao Tian, Xiaoliang Dai, Chih-Yao Ma, Zecheng He et al.CVPR 2023

Builds on3

  1. Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance AwarenessJeremiah Z. Liu, Zi Lin, Shreyas Padhy, Dustin Tran et al.NeurIPS 2020 · 604 citations
  2. Uncertainty Estimation Using a Single Deep Deterministic Neural NetworkJoost van Amersfoort, Lewis Smith, Yee Whye Teh, Yarin GalICML 2020 · 529 citations
  3. Angular Visual HardnessBeidi Chen, Weiyang Liu, Zhiding Yu, Jan Kautz et al.ICML 2020 · 57 citations

Related papers

  1. Geometric Logit Decoupling for Energy-Based Graph Out-of-distribution DetectionMin Wang, Hao Yang, Qing Cheng, Jincai HuangNeurIPS 2025
  2. RECON: Robust symmetry discovery via Explicit Canonical Orientation NormalizationAlonso Urbano, David Wilson Romero, Max Zimmer, Sebastian PokuttaICLR 2026 · 1 citation
  3. DecAug: Out-of-Distribution Generalization via Decomposed Feature Representation and Semantic AugmentationHaoyue Bai, Rui Sun, Lanqing Hong, Fengwei Zhou et al.AAAI 2021 · 88 citations
  4. Decompose-and-Compose: A Compositional Approach to Mitigating Spurious CorrelationFahimeh Hosseini Noohdani, Parsa Hosseini, Aryan Yazdan Parast, Hamidreza Yaghoubi Araghi et al.CVPR 2024 · 3 citations
  5. Simple and Effective Out-of-Distribution Detection via Cosine-based Softmax LossSoonCheol Noh, DongEon Jeong, Jee-Hyong LeeICCV 2023 · 6 citations
Dusk over the sea between two cliffs drawn in fine vertical lines