论文标题
理解依赖性：使用依赖度量的有效的黑盒解释
Making Sense of Dependence: Efficient Black-box Explanations Using Dependence Measure
论文作者
论文摘要
本文提出了一种基于Hilbert-Schmidt独立标准（HSIC）的新有效的黑盒归因方法，这是一种基于再现核Hilbert Space（RKHS）的依赖度度量。 HSIC测量了基于分布的内核的输入图像区域之间的依赖性和模型的输出。因此，它提供了由RKHS表示功能丰富的解释。可以非常有效地估计HSIC，与其他黑盒归因方法相比，大大降低了计算成本。我们的实验表明，HSIC的速度比以前最佳的Black-Box归因方法快8倍。确实，我们改善或匹配了黑盒和白色框归因方法的最新方法，用于具有各种最近的模型体系结构的Imagenet上的几个保真度指标。重要的是，我们表明这些进步可以被转移到有效而忠实地解释诸如Yolov4之类的对象检测模型。最后，我们通过提出一种新的内核来扩展传统的归因方法，从而实现基于HSIC的重要性分数的类似方差分析的正交分解，从而使我们不仅可以评估每个图像贴片的重要性，还可以评估其成对相互作用的重要性。我们的实施可在https://github.com/paulnovello/hsic-attribution-method上获得。
This paper presents a new efficient black-box attribution method based on Hilbert-Schmidt Independence Criterion (HSIC), a dependence measure based on Reproducing Kernel Hilbert Spaces (RKHS). HSIC measures the dependence between regions of an input image and the output of a model based on kernel embeddings of distributions. It thus provides explanations enriched by RKHS representation capabilities. HSIC can be estimated very efficiently, significantly reducing the computational cost compared to other black-box attribution methods. Our experiments show that HSIC is up to 8 times faster than the previous best black-box attribution methods while being as faithful. Indeed, we improve or match the state-of-the-art of both black-box and white-box attribution methods for several fidelity metrics on Imagenet with various recent model architectures. Importantly, we show that these advances can be transposed to efficiently and faithfully explain object detection models such as YOLOv4. Finally, we extend the traditional attribution methods by proposing a new kernel enabling an ANOVA-like orthogonal decomposition of importance scores based on HSIC, allowing us to evaluate not only the importance of each image patch but also the importance of their pairwise interactions. Our implementation is available at https://github.com/paulnovello/HSIC-Attribution-Method.
