Learning Robust Representations by Projecting Superficial Statistics Out
Haohan Wang Carnegie Mellon University Pittsburgh, PA USA、Zexue He Beijing Normal University Beijing, China、Zachary C. Lipton Carnegie Mellon University Pittsburgh, PA, USA、Eric P. Xing Carnegie Mellon University Pittsburgh, PA, USA
Abstract
Despite impressive performance as evaluated on i.i.d. holdout data, deep neural networks depend heavily on superficial statistics of the training data and are liable to break under distribution shift. For example, subtle changes to the background or texture of an image can break a seemingly powerful classifier. Building on previous work on domain generalization, we hope to produce a classifier that will generalize to previously unseen domains, even when domain identifiers are not available during training. This setting is challenging because the model may extract many distribution-specific (superficial) signals together with distribution-agnostic (semantic) signals. To overcome this challenge, we incorporate the gray-level co-occurrence matrix (GLCM) to extract patterns that our prior knowledge suggests are superficial: they are sensitive to texture but unable to capture the gestalt of an image. Then we introduce two techniques for improving our networks’ out-of-sample performance. The first method is built on the reverse gradient method that pushes our model to learn representations from which the GLCM representation is not predictable. The second method is built on the independen
中文速览
深度神经网络在真实场景中常因"捷径"——比如图片的纹理、背景等表面统计特征——而在训练域之外表现大幅下滑，如何让模型在完全没见过的新域上依然泛化，是领域泛化（Domain Generalization）的核心难题。针对这一问题，研究者引入了灰度共生矩阵（Gray-Level Co-occurrence Matrix, GLCM）的思路，设计了一个可微的神经网络模块（NGLCM），专门用来捕捉图像中的纹理这类"表面信息"，然后提出了两种让主干网络主动"遗忘"这些表面特征的方法：一是基于反向梯度的对抗训练（ADV），二是将主干网络的表示投影到与纹理表示正交的子空间（HEX），从而逼迫模型转而依赖更本质的语义信号。在多个标准领域泛化基准上的实验表明，该方法无需在训练时看到目标域的任何样本，性能却与那些需要目标域数据的方法相当甚至更优，说明"屏蔽已知表面特征"是提升模型跨域鲁棒性的一条切实可行的路径。
原文 arXiv:1903.06256;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1903.06256v1