Overconfident Prediction

Overconfident prediction in machine learning models, where predicted probabilities inaccurately reflect the model's true uncertainty, is a significant problem hindering the reliable deployment of AI systems. Current research focuses on improving model calibration through techniques like modifying loss functions (e.g., focal loss variants, entropy regularization), employing novel normalization methods (e.g., logit normalization, hyperspherical projections), and leveraging ensemble methods or Bayesian approaches to better quantify uncertainty. Addressing overconfidence is crucial for enhancing the trustworthiness and safety of AI in high-stakes applications, such as healthcare and autonomous driving, where accurate uncertainty estimation is paramount for responsible decision-making.

Papers

May 23, 2024

Unraveling overoptimism and publication bias in ML-driven science
Pouria Saidi, Gautam Dasarathy, Visar Berisha
Machine Learning Machine Learning Model Overconfident Prediction Learning Curve Publication Bias Reward Overoptimization

May 20, 2024

Beyond Calibration: Assessing the Probabilistic Fit of Neural Regressors via Conditional Congruence
Spencer Young, Cole Edgren, Riley Sinema, Andrew Hall, Nathan Dong, Porter Jenkins
High Dimensional Uncertainty Quantification Calibration Error Overconfident Prediction Probabilistic Neural Network Calibration Measure

February 13, 2024

Selective Learning: Towards Robust Calibration with Dynamic Regularization
Zongbo Han, Yifeng Yang, Changqing Zhang, Linjun Zhang, Joey Tianyi Zhou, Qinghua Hu, Huaxiu Yao
Calibrated Confidence Overconfident Prediction Robust Calibration Selective Learning

October 27, 2023

Classifier-head Informed Feature Masking and Prototype-based Logit Smoothing for Out-of-Distribution Detection
Zhuohao Sun, Yiqiao Qiu, Zhijun Tan, Weishi Zheng, Ruixuan Wang
Distribution Detection Distribution Data Feature Masking Overconfident Prediction Distribution Detection Benchmark

August 29, 2023

Uncertainty Aware Training to Improve Deep Learning Model Calibration for Classification of Cardiac MR Images
Tareen Dawood, Chen Chen, Baldeep S. Sidhua, Bram Ruijsink, Justin Goulda, Bradley Porter, Mark K. Elliott, Vishal Mehta, Christopher A. Rinaldi, Esther Puyol-Anton, Reza Razavi, Andrew P. King
Deep Learning Classification Code Cardiac Magnetic Resonance Overconfident Prediction Uncertainty Aware Training Cardiac Resynchronization Therapy

August 21, 2023

Overcoming Overconfidence for Active Learning
Yujin Hwang, Won Jo, Juyoung Hong, Yukyung Choi
Active Learning Effective Augmentation Overconfident Prediction High Confidence Prediction

July 24, 2023

Rethinking Data Distillation: Do Not Overlook Calibration
Dongyao Zhu, Bowen Lei, Jie Zhang, Yanbo Fang, Ruqi Zhang, Yiqun Xie, Dongkuan Xu
Calibration Performance Dataset Distillation Temperature Scaling Distilled Dataset Data Distillation Overconfident Prediction Masked Distillation

May 28, 2023

T2FNorm: Extremely Simple Scaled Train-time Feature Normalization for OOD Detection
Sudarshan Regmi, Bibek Panthi, Sakar Dotel, Prashnna K. Gyawali, Danail Stoyanov, Binod Bhattarai
Distribution Sample Understanding Out of Distribution Feature Normalization OOD Detection Overconfident Prediction

May 23, 2023

Dual Focal Loss for Calibration
Linwei Tao, Minjing Dong, Chang Xu
Calibration Performance Focal Loss Second Ranked Logits Overconfident Prediction Calibration Network

April 20, 2023

Learning Sample Difficulty from Pre-trained Models for Reliable Prediction
Peng Cui, Dan Zhang, Zhijie Deng, Yinpeng Dong, Jun Zhu
Pre Trained Model Large Scale Pre Trained Model Entropy Regularization Overconfident Prediction Reliable Prediction

December 1, 2022

Deep Kernel Learning for Mortality Prediction in the Face of Temporal Shift
Miguel Rios, Ameen Abu-Hanna
Human Face Neural Model Prediction Uncertainty Temporal Shift Neural Network Baseline Mortality Prediction Overconfident Prediction Deep Kernel Learning

November 29, 2022

Penalizing Confident Predictions on Largely Perturbed Inputs Does Not Improve Out-of-Distribution Generalization in Question Answering
Kazutoshi Shinoda, Saku Sugawara, Akiko Aizawa
Strong Generalization Question Answering Distribution Generalization Shuffle Model Low Confidence Input Perturbation Overconfident Prediction

October 23, 2022

On double-descent in uncertainty quantification in overparametrized models
Lucas Clarté, Bruno Loureiro, Florent Krzakala, Lenka Zdeborová
Full Model Uncertainty Quantification Double Descent Overparametrization Bound Overconfident Prediction Uncertainty Measure

June 23, 2022

WeShort: Out-of-distribution Detection With Weak Shortcut structure
Jinhong Lin
Neural Network Distribution Detection Understanding Out of Distribution Shortcut Pattern Avoidance Loss ImageNet Benchmark Overconfident Prediction

May 19, 2022

Mitigating Neural Network Overconfidence with Logit Normalization
Hongxin Wei, Renchunzi Xie, Hao Cheng, Lei Feng, Bo An, Yixuan Li
Overconfident Prediction Distribution Input Logit Calibration Confidence Difference

March 17, 2022

Confidence Calibration for Intent Detection via Hyperspherical Space and Rebalanced Accuracy-Uncertainty Loss
Yantao Gong, Cao Liu, Fan Yang, Xunliang Cai, Guanglu Wan, Jiansong Chen, Weipeng Zhang, Houfeng Wang
Intent Detection Confidence Calibration Overconfident Prediction Accuracy Loss HyperSpherical Energy

February 16, 2022

Reducing Overconfidence Predictions for Autonomous Driving Perception
Gledson Melotti, Cristiano Premebida, Jordan J. Bird, Diego R. Faria, Nuno Gonçalves
Autonomous Driving Object Recognition State of the Art Deep Pre Trained Network Sigmoid Function Overconfident Prediction Driving Perception

December 15, 2021

Taming Overconfident Prediction on Unlabeled Data from Hindsight
Jing Li, Yuangang Pan, Ivor W. Tsang
Semi Supervised Learning Unlabeled Data Prediction Uncertainty Uncertainty Prediction Hindsight Instruction Overconfident Prediction Distillation Learning

November 25, 2021

Reliable Graph Neural Networks for Drug Discovery Under Distributional Shift
Kehang Han, Balaji Lakshminarayanan, Jeremiah Liu
Graph Neural Network Distribution Shift Drug Discovery GNN Cl Overconfident Prediction Trustworthy Graph

November 5, 2021

Mixtures of Laplace Approximations for Improved Post-Hoc Uncertainty in Deep Learning
Runa Eschenhagen, Erik Daxberger, Philipp Hennig, Agustinus Kristiadi
Deep Learning Deep Neural Network Mixture Component Bayesian Neural Network Gaussian Mixture Model Deep Ensemble Overconfident Prediction Laplace Approximation