Hiroki Naganuma

LogitNorm for Calibration

Topic	LogitNorm for Calibration
Collaborator	LinkedIn
Venue	US Patent?
Status	WIP(Writing)

Schedule

TODO

Calibration 関連に向けて

実装

• OOD 向け手法の調査と実装
  • IRM を DeepCTR に実装する
    • FB IRM
    • IRM for NLI
    • DomainBed
    • Minimal Implementation of IRM in PyTorch
  • fishr とか OOD 向けの学習法を試す github

実験 PLot

• 実験やってる感を出す plot の script 試す
  • BS 比較もするか？
  • RCT のドキュメントどうにかする (これが一番大事かも)
    • Paper Survey
      • Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift
      • Assessing Generalization of SGD via Disagreement
      • Deep Learning Is Singular, and That’s Good
      • Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration
      • Mix-n-Match: Ensemble and Compositional Methods for Uncertainty Calibration in Deep Learning
      • Post-hoc Uncertainty Calibration for Domain Drift Scenarios
      • Training calibration-based counterfactual explainers for deep learning models in medical image analysis
      • Field-aware Calibration: A Simple and Empirically Strong Method for Reliable Probabilistic Predictions
      • Towards Trustworthy Predictions from Deep Neural Networks with Fast Adversarial Calibration
      • Improving Uncertainty Calibration of Deep Neural Networks via Truth Discovery and Geometric Optimization