arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.18992cs.CVcs.AIcs.LG
收藏

半监督概念瓶颈模型

Semi-supervised Concept Bottleneck Models

University of Virginia(弗吉尼亚大学);KTH Royal Institute of Technology(瑞典皇家理工学院);University of Georgia(佐治亚大学);Shanghai Jiao Tong University(上海交通大学);HKUST(香港科技大学);Ant Group(蚂蚁集团)

Lijie Hu, Tianhao Huang, Huanyi Xie, Xilin Gong, Chenyang Ren, Zhengyu Hu, Lu Yu, Ping Ma, Di Wang

AI总结:

针对概念瓶颈模型依赖大量专家标注概念、概念与输入显著性图易错位的问题,提出半监督概念瓶颈模型SSCBM,通过伪标签策略、对齐损失及有/无标注数据联合训练实现概念层面对齐,仅用10%标注数据性能接近全监督最优基线。

摘要:

概念瓶颈模型(Concept Bottleneck Models, CBMs)因能够为黑盒深度学习模型提供基于概念的解释,同时利用类人概念实现较高的最终预测准确率,受到了越来越多的关注。然而,当前CBMs的训练高度依赖数据集中标注概念的精度与丰富度。这些概念标签通常由专家提供,成本高昂,且需要投入大量资源与精力。此外,概念显著性图常与输入显著性图错位,导致概念预测对应到不相关的输入特征——这是一个与标注对齐相关的问题。为解决这些局限,我们提出了名为SSCBM(半监督概念瓶颈模型)的新框架。我们的SSCBM适用于标注数据稀缺的实际场景,通过在有标注和无标注数据上进行联合训练,并在概念层面对齐无标注数据,有效解决了上述问题。我们提出了一种生成伪标签的策略和一种对齐损失。实验表明,我们的SSCBM兼具有效性与高效性:仅使用10%的标注数据时,我们的模型在四个数据集上的概念准确率和任务准确率,与全监督学习设置下的最优基线相比,分别仅低2.44%和3.93%。

英文摘要:

Concept Bottleneck Models (CBMs) have garnered increasing attention due to their ability to provide concept-based explanations for black-box deep learning models while achieving high final prediction accuracy using human-like concepts. However, the training of current CBMs is heavily dependent on the precision and richness of the annotated concepts in the dataset. These concept labels are typically provided by experts, which can be costly and require significant resources and effort. Additionally, concept saliency maps frequently misalign with input saliency maps, causing concept predictions to correspond to irrelevant input features - an issue related to annotation alignment. To address these limitations, we propose a new framework called SSCBM (Semi-supervised Concept Bottleneck Model). Our SSCBM is suitable for practical situations where annotated data is scarce. By leveraging joint training on both labeled and unlabeled data and aligning the unlabeled data at the concept level, we effectively solve these issues. We proposed a strategy to generate pseudo labels and an alignment loss. Experiments demonstrate that our SSCBM is both effective and efficient. With only 10% labeled data, our model's concept and task accuracy on average across four datasets is only 2.44% and 3.93% lower, respectively, compared to the best baseline in the fully supervised learning setting.

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。

↑