Influence-Directed Explanations for Deep Convolutional Networks
Influence-Directed Explanations for Deep Convolutional Networks
复制标题
DOI：
10.1109/test.2018.8624792
复制
发表时间：
2018-02
期刊：
影响因子：
--
通讯作者：
Klas Leino;Linyi Li;S. Sen;Anupam Datta;Matt Fredrikson
中科院分区：
文献类型：
--
作者：
Klas Leino;Linyi Li;S. Sen;Anupam Datta;Matt Fredrikson
我们研究解释深度神经网络的一类丰富行为特性的问题。独特的是，我们的基于影响导向的解释方法通过深入网络内部来解决这个问题，即使用一种经公理化证明合理的影响度量来识别对感兴趣的量和分布具有高影响力的神经元，然后对这些神经元所代表的概念进行解释。我们通过在ImageNet上训练的卷积神经网络展示其多项独特能力来评估我们的方法。我们的评估表明，基于影响导向的解释（1）识别出能在实例间通用的有影响力的概念；（2）可用于提取网络关于某一类所学内容的“本质”；（3）分离出网络用于做决策和区分相关类别的单个特征。
We study the problem of explaining a rich class of behavioral properties of deep neural networks. Distinctively, our influence-directed explanations approach this problem by peering inside the network to identify neurons with high influence on a quantity and distribution of interest, using an axiomatically-justified influence measure, and then providing an interpretation for the concepts these neurons represent. We evaluate our approach by demonstrating a number of its unique capabilities on convolutional neural networks trained on ImageNet. Our evaluation demonstrates that influence-directed explanations (1) identify influential concepts that generalize across instances, (2) can be used to extract the “essence” of what the network learned about a class, and (3) isolate individual features the network uses to make decisions and distinguish related classes.