Multiple metric learning via local metric fusion

Adaptive distance metric learning based on the characteristics of data can significantly improve the learner's performance. Due to the limitations of single metric learning for heterogeneous data, multiple local metric learning has become an essential representative tool to describe local properties of data. Most existing multiple metric learning algorithms need to perform metric learning on a pre-obtained instance division. However, the number of clusters in the pre-obtained division affects the effectiveness of metric learning. To tackle this problem, we propose a Multiple Metric Learning via Local Metric Fusion (MML-LMF) framework, which unifies local metric learning and fusion of similar local met-rics into one metric and adaptively determines the number of local metrics. As an applica-tion of the MML-LMF framework to pairwise constraints, we devise a MML-LMF algorithm by constructing a concrete optimization model and acquiring a closed-form solution to the model. The experimental results on several benchmarks, person re-identification, and face verification datasets show that the performance of the proposed algorithm is superior to that of the existing state-of-the-art global and multiple metric learning algorithms.(c) 2022 Elsevier Inc. All rights reserved.

Automated summary

Adaptive distance metric learning based on data characteristics can improve learner's performance significantly. Multiple local metric learning is an essential tool to describe local properties of heterogeneous data. The proposed Multiple Metric Learning via Local Metric Fusion (MML-LMF) framework unifies local metric learning and fusion of similar local metrics, adaptively determines the number of local metrics, and outperforms existing state-of-the-art global and multiple metric learning algorithms, as demonstrated on various benchmarks and datasets.