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 2107.02575cs.CV
收藏

基于TupleInfoNCE的对比多模态融合

Contrastive Multimodal Fusion with TupleInfoNCE

Tsinghua University(清华大学);Google Research(谷歌研究院);Stanford University(斯坦福大学);UC Berkeley(加州大学伯克利分校);Peking University(北京大学);Xidian University(西安电子科技大学);Peng Cheng Laboratory(鹏城实验室)

Yunze Liu, Qingnan Fan, Shanghang Zhang, Hao Dong, Thomas Funkhouser, Li Yi

摘要:

提出了一种利用对比损失的多模态数据表示学习方法。传统的方法是对比不同的模式来学习它们之间共享的信息。然而,这种做法可能无法学习可能对下游任务有用的各种模式之间的互补协同作用。另一种方法是将所有模态连接到元组中,然后对比正元组和负元组对应。然而,这种做法可能只考虑较强的模式,而忽略较弱的模式。为了解决这些问题,我们提出了一个新的对比学习目标,TupleInfoNCE。它不仅基于正对应和负对应来对比元组,而且还通过使用描述不同场景的模态来组成新的负元组来对比元组。使用这些额外的否定项进行训练可以鼓励学习模型检查同一元组中模态之间的对应关系,确保不忽略弱模态。我们提供了一个基于互信息的理论证明,为什么这种方法的工作原理,我们提出了一个样本优化算法来生成积极和消极的样本,以最大限度地提高训练效果。我们发现TupleInfoNCE在三个不同的下游任务上显著优于以前的技术水平。

英文摘要:

This paper proposes a method for representation learning of multimodal data using contrastive losses. A traditional approach is to contrast different modalities to learn the information shared between them. However, that approach could fail to learn the complementary synergies between modalities that might be useful for downstream tasks. Another approach is to concatenate all the modalities into a tuple and then contrast positive and negative tuple correspondences. However, that approach could consider only the stronger modalities while ignoring the weaker ones. To address these issues, we propose a novel contrastive learning objective, TupleInfoNCE. It contrasts tuples based not only on positive and negative correspondences but also by composing new negative tuples using modalities describing different scenes. Training with these additional negatives encourages the learning model to examine the correspondences among modalities in the same tuple, ensuring that weak modalities are not ignored. We provide a theoretical justification based on mutual information for why this approach works, and we propose a sample optimization algorithm to generate positive and negative samples to maximize training efficacy. We find that TupleInfoNCE significantly outperforms the previous state of the arts on three different downstream tasks.

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。

↑