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 2212.07430cs.LGcs.AI
收藏

交互式概念瓶颈模型

Interactive Concept Bottleneck Models

Google Research(谷歌研究院)

Kushal Chauhan, Rishabh Tiwari, Jan Freyberg, Pradeep Shenoy, Krishnamurthy Dvijotham

AI总结:

本文提出交互式概念瓶颈模型,通过结合概念不确定性与预测影响的策略向人类查询概念标签,仅5次交互即在多个数据集上较基线提升5-10%准确率。

摘要:

概念瓶颈模型(CBMs)是一类可解释神经网络,首先预测与预测任务相关的人类可解释概念的标签,然后基于概念标签预测来预测最终标签。我们将 CBMs 扩展到交互式预测设置,在该设置中,模型可以向人类协作者查询某些概念的标签。我们开发了一种交互策略,在预测时选择请求哪些概念的标签,以最大程度地改善最终预测。我们证明,一种结合概念预测不确定性与该概念对最终预测影响的简单策略取得了强劲性能,并优于静态方法以及文献中提出的主动特征获取方法。我们表明,在 Caltech-UCSD Birds、CheXpert 和 OAI 数据集上,仅需 5 次交互,交互式 CBM 相比具有竞争力的基线即可实现 5-10% 的准确率提升。

英文摘要:

Concept bottleneck models (CBMs) are interpretable neural networks that first predict labels for human-interpretable concepts relevant to the prediction task, and then predict the final label based on the concept label predictions. We extend CBMs to interactive prediction settings where the model can query a human collaborator for the label to some concepts. We develop an interaction policy that, at prediction time, chooses which concepts to request a label for so as to maximally improve the final prediction. We demonstrate that a simple policy combining concept prediction uncertainty and influence of the concept on the final prediction achieves strong performance and outperforms static approaches as well as active feature acquisition methods proposed in the literature. We show that the interactive CBM can achieve accuracy gains of 5-10% with only 5 interactions over competitive baselines on the Caltech-UCSD Birds, CheXpert and OAI datasets.

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。

↑