论文标题
ERIC：从卷积中推断的关系
ERIC: Extracting Relations Inferred from Convolutions
论文作者
论文摘要
我们的主要贡献是表明，可以使用逻辑程序近似卷积神经网络多层的内核行为。提取的逻辑程序产生与原始模型相关的精确度，尽管尤其是一定的信息丢失，因为将多层的近似值链在一起，或者随着较低的层被定量。我们还表明，提取的程序可以用作进一步理解CNN行为的框架。具体而言，它可用于识别值得更深入检查的关键内核，并以逻辑规则的形式识别与其他内核的关系。最后，我们对我们从最后一个卷积层中提取的规则进行初步定性评估，并表明所识别的内核是象征性的，因为它们对有效分配输出类别的类似图像的集合对具有独特特征的子类反应。
Our main contribution is to show that the behaviour of kernels across multiple layers of a convolutional neural network can be approximated using a logic program. The extracted logic programs yield accuracies that correlate with those of the original model, though with some information loss in particular as approximations of multiple layers are chained together or as lower layers are quantised. We also show that an extracted program can be used as a framework for further understanding the behaviour of CNNs. Specifically, it can be used to identify key kernels worthy of deeper inspection and also identify relationships with other kernels in the form of the logical rules. Finally, we make a preliminary, qualitative assessment of rules we extract from the last convolutional layer and show that kernels identified are symbolic in that they react strongly to sets of similar images that effectively divide output classes into sub-classes with distinct characteristics.
