提出可解释的极小极大神经网络，单个神经元决定输出，便于追踪决策过程。
On the explainability of max-plus neural networks
- 通过极小极大结构设计，初始化时等价于无穷范数下的k-medoids聚类
- 训练后仅一个最激活神经元决定输出值，实现决策路径可追溯
- 提出像素脆弱性度量，对肺炎影像分类优于SHAP和积分梯度
我们研究了最近提出的线性-极小-极大神经网络的可解释性。在初始化时，该模型可解释为使用无穷范数的k-medoids聚类。随后，通过次梯度下降进行训练以更好地拟合数据。尽管该模型已被证明是通用逼近器，但其决策过程仍可追踪，因为输出值仅由单一最激活神经元决定。基于此特性，我们设计了一种像素脆弱性度量，用于判断单个像素的变化是否可能引发分类结果改变。在PneumoniaMnist数据集上的实验表明，该方法生成的解释相较于SHAP和积分梯度更具优势。
原文摘要 · Abstract (English)
We investigate the explanability properties of the recently proposed linear-min-max neural networks. At initialization, they can be interpreted as k-medoids with the infinity norm as a distance. Then, they are trained using subgradient descent to better fit the data. The model has been shown to be a universal approximator. Yet, we can trace the decision process because a single most activated neuron is responsible for the value of the output. Using this property, we designed a pixel fragility measure that determines whether changes to a single pixel may be responsible to a change in the classification output. Experiments on the PneumoniaMnist dataset show that this explanation for the output of the neural network compares favorably to SHAP and Integrated Gradient.
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。