Machine learningTrustworthy ML
模型校准
模型校准是一种事后技术，用于调整已训练分类器的概率输出，使预测置信度得分与经验结果频率相匹配。当对所有置信度为 p 的预测中，恰好有 p 的比例是正确的时，则称分类器是完美校准的。Guo 等人 (2017) 严谨地记录了现代深度神经网络系统性的失校准问题，他们表明使用标准交叉熵损失训练的网络倾向于过度自信，并提出温度缩放作为一种简单有效的补救方法。
阅读完整方法
仅限会员
登录使用免费账户登录即可阅读本节。
Method map
The neighbourhood of related methods — select a node to explore.
来源
- Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On calibration of modern neural networks. International Conference on Machine Learning, 1321–1330. link ↗
如何引用本页
ScholarGate. (2026, June 2). Probability Calibration of Classifiers. ScholarGate. https://scholargate.app/zh/machine-learning/model-calibration
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
Compare side by side →