用贝叶斯方法改进神经网络概率输出，让预测更可信。
The Bayesian Confidence (BACON) Estimator for Deep Neural Networks
- 基于验证数据构建几何模型，用贝叶斯法则估算输出概率。
- 在CIFAR-10上，ResNet-18和EfficientNet-B0的ECE和ACE误差更低。
- 对类别不平衡数据，能更好校准真实分布下的概率估计。
本文提出深度神经网络的贝叶斯置信度估计器（BACON）。当前将Softmax输出视为类概率的做法易产生极端预测。本文扩展Waagen的几何建模方法，利用验证数据通过贝叶斯规则估算输出向量的概率，得到似然与归一化值。该方法在CIFAR-10数据集的不平衡测试集上，于ResNet-18（85%准确率）和EfficientNet-B0（95%准确率）上均优于Softmax，显著降低ECE与ACE校准误差，仅在极高准确率边缘情况例外。此外，在使用实际类别比例时，采用ACE指标的BACON在不平衡测试集上表现出更优的概率校准能力。
原文摘要 · Abstract (English)
This paper introduces the Bayesian Confidence Estimator (BACON) for deep neural networks. Current practice of interpreting Softmax values in the output layer as probabilities of outcomes is prone to extreme predictions of class probability. In this work we extend Waagen's method of representing the terminal layers with a geometric model, where the probability associated with an output vector is estimated with Bayes' Rule using validation data to provide likelihood and normalization values. This estimator provides superior ECE and ACE calibration error compared to Softmax for ResNet-18 at 85% network accuracy, and EfficientNet-B0 at 95% network accuracy, on the CIFAR-10 dataset with an imbalanced test set, except for very high accuracy edge cases. In addition, when using the ACE metric, BACON demonstrated improved calibration error when estimating probabilities for the imbalanced test set when using actual class distribution fractions.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。