文件列表:
VISION DIFFMASK:可微分 patch 掩模的视觉 transformer 忠实解释【英文版】.pdf |
下载文档 |
资源简介
>
英文标题:VISION DIFFMASK: Faithful Interpretation of Vision Transformers with Differentiable Patch Masking中文摘要:提出了一种名为 VISION DIFFMASK 的后处理可解释性方法,使用模型隐藏层的激活来预测对其最终预测有贡献的输入的相关部分,并使用门控机制识别原始输入的最小子集,以维持对类的预测分布的预测结果;结果表明该方法在 CIFAR-10 和 ImageNet-1K 上比其他最先进的归因方法更为可靠。英文摘要:The lack of interpretability of the Vision Transformer may hinder its use incritical real-world applications despite its effectiveness. To overcome thisissue, we propose a post-hoc interpretability method called VISION DIFFMASK,which use
加载中...
已阅读到文档的结尾了