arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2007.08194cs.CVcs.LG
收藏

区分类特定滤波器训练可解释卷积神经网络

Training Interpretable Convolutional Neural Networks by Differentiating Class-specific Filters

Tsinghua University(清华大学);Tsinghua SIGS(清华大学深圳国际研究生院);Peng Cheng Lab(鹏城实验室);ByteDance AI Lab(字节跳动人工智能实验室);University of Southern California(南加州大学)

Haoyu Liang, Zhihao Ouyang, Yuyuan Zeng, Hang Su, Zihao He, Shu-Tao Xia, Jun Zhu, Bo Zhang

摘要:

卷积神经网络(CNN)已成功地用于一系列任务。然而,CNN通常被视为“黑箱”,缺乏可解释性。一个主要原因是由于过滤器类纠缠-过滤器和类之间复杂的多对多对应关系。大多数现有的作品尝试在预先训练的模型上进行事后解释,而忽略了减少模型背后的纠缠。相比之下,我们专注于在训练过程中减轻过滤器类纠缠。受细胞分化的启发,我们提出了一种新的策略,通过鼓励类特定的过滤器来训练可解释的CNN,其中每个过滤器只响应一个(或几个)类。具体来说,我们设计了一个可学习的稀疏类特定门(CSG)结构,以灵活的方式为每个过滤器分配一个(或几个)类。该门只允许在输入样本来自特定类时通过过滤器的激活。大量的实验证明了我们的方法在生成稀疏和高度类相关的输入表示方面的出色性能,这导致了更强的可解释性。此外,与标准训练策略相比,我们的模型在对象定位和对抗性样本检测等应用中显示出优势。代码链接:https://github.com/hyliang96/CSGCNN.

英文摘要:

Convolutional neural networks (CNNs) have been successfully used in a range of tasks. However, CNNs are often viewed as "black-box" and lack of interpretability. One main reason is due to the filter-class entanglement -- an intricate many-to-many correspondence between filters and classes. Most existing works attempt post-hoc interpretation on a pre-trained model, while neglecting to reduce the entanglement underlying the model. In contrast, we focus on alleviating filter-class entanglement during training. Inspired by cellular differentiation, we propose a novel strategy to train interpretable CNNs by encouraging class-specific filters, among which each filter responds to only one (or few) class. Concretely, we design a learnable sparse Class-Specific Gate (CSG) structure to assign each filter with one (or few) class in a flexible way. The gate allows a filter's activation to pass only when the input samples come from the specific class. Extensive experiments demonstrate the fabulous performance of our method in generating a sparse and highly class-related representation of the input, which leads to stronger interpretability. Moreover, comparing with the standard training strategy, our model displays benefits in applications like object localization and adversarial sample detection. Code link: https://github.com/hyliang96/CSGCNN.

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。

↑