少即是多：发现简洁的网络解释
Less is More: Discovering Concise Network Explanations
AI总结：
本文提出 DCNE 方法，基于 CRP 自动筛选少量、多样且可解释的关键视觉概念，在鸟类分类数据集上以仅 1/30 的解释数量实现接近 CRP 的质量。
摘要：
我们提出了 Discovering Conceptual Network Explanations（DCNE，发现概念性网络解释），这是一种生成人类可理解的视觉解释以增强深度神经图像分类器可解释性的新方法。我们的方法会自动寻找对于区分类别至关重要的视觉解释。这是通过同时优化三个准则实现的：解释应当数量少、具有多样性且可被人类解释。我们的方法建立在最近提出的 Concept Relevance Propagation（CRP，概念相关性传播）可解释性方法之上。尽管 CRP 能有效描述单个神经元激活，但它会生成过多概念，从而影响人类理解。与之不同，DCNE 会选择少数最重要的解释。我们引入了一个新的评估数据集，其核心是具有挑战性的鸟类分类任务，使我们能够比较 DCNE 的解释与人类专家定义的解释之间的一致性。与现有 eXplainable Artificial Intelligence（XAI，可解释人工智能）方法相比，DCNE 在概括网络解释时，在简洁性和完整性之间取得了理想的权衡。它生成的解释数量仅为 CRP 的 1/30，而解释质量仅有轻微下降。DCNE 代表了让神经网络决策对人类而言更易理解、更可解释的一步进展，为 XAI 和模型对齐领域的研究者与从业者提供了有价值的工具。
英文摘要：
We introduce Discovering Conceptual Network Explanations (DCNE), a new approach for generating human-comprehensible visual explanations to enhance the interpretability of deep neural image classifiers. Our method automatically finds visual explanations that are critical for discriminating between classes. This is achieved by simultaneously optimizing three criteria: the explanations should be few, diverse, and human-interpretable. Our approach builds on the recently introduced Concept Relevance Propagation (CRP) explainability method. While CRP is effective at describing individual neuronal activations, it generates too many concepts, which impacts human comprehension. Instead, DCNE selects the few most important explanations. We introduce a new evaluation dataset centered on the challenging task of classifying birds, enabling us to compare the alignment of DCNE's explanations to those of human expert-defined ones. Compared to existing eXplainable Artificial Intelligence (XAI) methods, DCNE has a desirable trade-off between conciseness and completeness when summarizing network explanations. It produces 1/30 of CRP's explanations while only resulting in a slight reduction in explanation quality. DCNE represents a step forward in making neural network decisions accessible and interpretable to humans, providing a valuable tool for both researchers and practitioners in XAI and model alignment.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
