Evaluating model calibration in classification

Abstract

Probabilistic classifiers output a probability distribution on target classes rather than just a class prediction. Besides providing a clear separation of prediction and decision making, the main advantage of probabilistic models is their ability to represent uncertainty about predictions. In safety-critical applications, it is pivotal for a model to possess an adequate sense of uncertainty, which for probabilistic classifiers translates into outputting probability distributions that are consistent with the empirical frequencies observed from realized outcomes. A classifier with such a property is called calibrated. In this work, we develop a general theoretical calibration evaluation framework grounded in probability theory, and point out subtleties present in model calibration evaluation that lead to refined interpretations of existing evaluation techniques. Lastly, we propose new ways to quantify and visualize miscalibration in probabilistic classification, including novel multidimensional reliability diagrams.

Cited by in corpus (35)

• A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges
• Classifier Calibration: A survey on how to assess and improve predicted class probabilities
• Measuring Calibration in Deep Learning
• Verified Uncertainty Calibration
• Calibrating Deep Neural Networks using Focal Loss
• Revisiting the Calibration of Modern Neural Networks
• Danish Fungi 2020 -- Not Just Another Image Recognition Dataset
• Calibrating AI Models for Few-Shot Demodulation via Conformal Prediction
• Assessing Generalization of SGD via Disagreement
• Can We Leverage Predictive Uncertainty to Detect Dataset Shift and Adversarial Examples in Android Malware Detection?
• Multi-Class Uncertainty Calibration via Mutual Information Maximization-based Binning
• Calibration of Model Uncertainty for Dropout Variational Inference
• A Unified View of Label Shift Estimation
• On Calibration and Out-of-domain Generalization
• Should Ensemble Members Be Calibrated?
• Combining Ensembles and Data Augmentation can Harm your Calibration
• Distribution-free binary classification: prediction sets, confidence intervals and calibration
• Revisiting the Evaluation of Uncertainty Estimation and Its Application to Explore Model Complexity-Uncertainty Trade-Off
• Distribution-free uncertainty quantification for classification under label shift
• Are Graph Neural Networks Miscalibrated?
• Confidence-based Estimators for Predictive Performance in Model Monitoring
• INVIGORATE: Interactive Visual Grounding and Grasping in Clutter
• Calibration of Neural Networks using Splines
• Robust quantum dots charge autotuning using neural network uncertainty
• On the Usefulness of the Fit-on-the-Test View on Evaluating Calibration of Classifiers
• Top-label calibration and multiclass-to-binary reductions
• Plots of the cumulative differences between observed and expected values of ordered Bernoulli variates
• Robustness and Reliability When Training With Noisy Labels
• Explaining generalization in deep learning: progress and fundamental limits
• Point & Grasp: Flexible Selection of Out-of-Reach Objects Through Probabilistic Cue Integration
• Calibration tests beyond classification
• Diagnostic Uncertainty Calibration: Towards Reliable Machine Predictions in Medical Domain
• Identifying and Exploiting Structures for Reliable Deep Learning
• Prostate Tissue Grading with Deep Quantum Measurement Ordinal Regression
• Calibrating Structured Output Predictors for Natural Language Processing