[TPAMI-2023] Towards Improved and Interpretable Deep Metric Learning via Attentive Grouping

论文阅读 [TPAMI-2023] Towards Improved and Interpretable Deep Metric Learning via Attentive Grouping

论文搜索(studyai.com)

搜索论文: Towards Improved and Interpretable Deep Metric Learning via Attentive Grouping

搜索论文: http://www.studyai.com/search/whole-site/?q=Towards+Improved+and+Interpretable+Deep+Metric+Learning+via+Attentive+Grouping&fr=csdn

关键字(Keywords)

Measurement; Training; Semantics; Testing; Task analysis; Convolutional neural networks; Tensors; Deep metric learning; grouping; attention; interpretability; invariance

机器学习; 自然语言处理; 运筹与优化

(深度)度量学习; 注意力机制; 语义分析; 语义特征; 损失函数

摘要(Abstract)

Grouping has been commonly used in deep metric learning for computing diverse features.

分组通常用于深度度量学习，以计算不同的特征。

To improve the performance and interpretability, we propose an improved and interpretable grouping method to be integrated flexibly with any metric learning framework.

为了提高性能和可解释性，我们提出了一种改进的、可解释的分组方法，可与任何度量学习框架灵活集成。

Specifically, our method is based on the attention mechanism with a learnable query for each group.

具体来说，我们的方法基于注意力机制，每个组都有一个可学习的查询。

The query is fully trainable and can capture group-specific information when combined with the diversity loss.

该查询是完全可训练的，当与多样性损失相结合时，可以捕获特定于组的信息。

An appealing property of our method is that it naturally lends itself interpretability.

我们方法的一个吸引人的特性是，它自然具有可解释性。

The attention scores between the learnable query and each spatial position can be interpreted as the importance of that position.

可学习查询和每个空间位置之间的注意力得分可以解释为该位置的重要性。

We formally show that our proposed grouping method is invariant to spatial permutations of features.

我们正式证明了我们提出的分组方法对特征的空间排列是不变的。

When used as a module in convolutional neural networks, our method leads to translational invariance.

当在卷积神经网络中用作模块时，我们的方法导致平移不变性。

We conduct comprehensive experiments to evaluate our method.

我们进行了全面的实验来评估我们的方法。

Our quantitative results indicate that the proposed method outperforms prior methods consistently and significantly across different datasets, evaluation metrics, base models, and loss functions.

我们的定量结果表明，所提出的方法在不同的数据集、评估指标、基础模型和损失函数上一致且显著地优于现有方法。

For the first time to the best of our knowledge, our interpretation results clearly demonstrate that the proposed method enables the learning of diverse and stable semantic features across groups…

据我们所知，我们的解释结果首次清楚地表明，所提出的方法能够跨组学习多样和稳定的语义特征。

作者(Authors)

[‘Xinyi Xu’, ‘Zhengyang Wang’, ‘Cheng Deng’, ‘Hao Yuan’, ‘Shuiwang Ji’]