On Calibration of Modern Neural Networks

Confidence calibration – the problem of predicting probability estimates representative of the true correctness likelihood – is important for classification models in many applications. We discover...

On Calibration of Modern Neural Networks Chuan Guo, Geoff Pleiss, Yu Sun, Kilian Q. Weinberger Proceedings of the 34th International Conference on Machine Learning , PMLR 70:1321-1330, 2017. Abstract Confidence calibration – the problem of predicting probability estimates representative of the true correctness likelihood – is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important facto

related reading

• [1706.04599] On Calibration of Modern Neural Networksarxiv.org
• On Calibration of Modern Neural Networksarxiv.org
• guo17a.pdfproceedings.mlr.press
• Understanding deep learning requires rethinking generalizationarxiv.org
• On Getting Confidence Estimates from Neural Networks | Bharath's notesbharathpbhat.github.io
• nn-notes.pdfboris-hanin.github.io
• arxiv.org/pdf/1805.08522arxiv.org
• Infinite Limits of Neural Networks - Kempner Institutekempnerinstitute.harvard.edu
• [2605.01172] A Theory of Generalization in Deep Learningarxiv.org
• [2201.02177] Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasetsarxiv.org
• Neural networks: from the perceptron to deep netsarxiv.org
• Clare Lyle | The state of plasticity in 2025clarelyle.com