隐藏的单调性：通过 DC 分解解释深度神经网络
Hidden Monotonicity: Explaining Deep Neural Networks via their DC Decomposition
把训练好的 ReLU 网络拆成两个单调凸部分（DC 分解）并解决权重爆炸的数值问题，可直接替换现有 CAM/LRP 的传播规则，得到更忠实的显著性图；同时可用「两个单调网络之差」的架构训练天生自解释的视觉模型。
原文依据 we use an adaptation of the decomposition of a trained ReLU network into two monotone and convex parts
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论已有多种场景表明，单调性能够提升神经网络的可解释性。然而，并非每个函数都能被单调神经网络很好地逼近。我们证明，单调性仍可通过两种方式用于提升可解释性。首先，我们采用一种改编方法，将训练好的 ReLU 网络分解为两个单调且凸的部分，从而克服该过程中权重固有爆炸所带来的数值障碍。我们提出的显著性方法 SplitCAM 与 SplitLRP 在 ImageNet-S 上，于 VGG16 和 ResNet18 两个网络上，在全部 Quantus 显著性指标类别上均优于当前最优结果。其次，我们证明，将模型训练为两个单调神经网络之差，可以得到一个具有强自解释性质的系统。
英文原文摘要 839 字符
It has been demonstrated in various contexts that monotonicity leads to better explainability in neural networks. However, not every function can be well approximated by a monotone neural network.We demonstrate that monotonicity can still be used in two ways to boost explainability. First, we use an adaptation of the decomposition of a trained ReLU network into two monotone and convex parts, thereby overcoming numerical obstacles from an inherent blowup of the weights in this procedure. Our proposed saliency methods -- SplitCAM and SplitLRP --improve onstate of the art results on both VGG16 and Resnet18 networks on ImageNet-S across all Quantus saliency metric categories.Second, we exhibit that training a model as the difference between two monotone neural networks results in a system with strong self-explainability properties.