面向基于梯度的解释的流形假设
The Manifold Hypothesis for Gradient-Based Explanations
AI总结：
本研究提出基于梯度的解释算法的感知一致性与特征归因和数据流形切空间的对齐度相关的假设，通过多数据集实验验证该假设，并发现主流事后方法、对抗训练可提升对齐度，为解释算法优化提供方向。
摘要：
基于梯度的解释算法何时能提供与感知一致的解释？我们提出一项准则：特征归因需要与数据流形的切空间对齐。为给该假设提供证据，我们引入一个基于变分自编码器（variational autoencoders）的框架，可用于估计并生成图像流形。通过在MNIST、EMNIST、CIFAR10、X射线肺炎检测及糖尿病视网膜病变检测等一系列不同数据集上开展实验，我们证明：特征归因与数据切空间的对齐程度越高，其往往就越符合感知一致性。我们随后发现，Integrated Gradients、SmoothGrad等主流事后方法给出的归因，与数据流形的对齐程度远强于原始梯度。对抗训练也能提升模型梯度与数据流形的对齐度。据此我们提出，解释算法应主动促使其解释结果与数据流形对齐。本文是CVPR研讨会论文的扩展版本，代码可在https://github.com/tml-tuebingen/explanations-manifold获取。
英文摘要：
When do gradient-based explanation algorithms provide perceptually-aligned explanations? We propose a criterion: the feature attributions need to be aligned with the tangent space of the data manifold. To provide evidence for this hypothesis, we introduce a framework based on variational autoencoders that allows to estimate and generate image manifolds. Through experiments across a range of different datasets -- MNIST, EMNIST, CIFAR10, X-ray pneumonia and Diabetic Retinopathy detection -- we demonstrate that the more a feature attribution is aligned with the tangent space of the data, the more perceptually-aligned it tends to be. We then show that the attributions provided by popular post-hoc methods such as Integrated Gradients and SmoothGrad are more strongly aligned with the data manifold than the raw gradient. Adversarial training also improves the alignment of model gradients with the data manifold. As a consequence, we suggest that explanation algorithms should actively strive to align their explanations with the data manifold. This is an extended version of a CVPR Workshop paper. Code is available at https://github.com/tml-tuebingen/explanations-manifold.
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。
