Improving Generalization of Metric Learning via Listwise Self-distillation
This work addresses overfitting issues in metric learning for applications like image retrieval, though it is incremental as it builds on existing frameworks.
The paper tackles overfitting in deep metric learning by proposing Listwise Self-Distillation (LSD), a regularization method that adaptively assigns distance targets to sample pairs, resulting in improved generalization and consistent performance boosts across various methods and datasets.
Most deep metric learning (DML) methods employ a strategy that forces all positive samples to be close in the embedding space while keeping them away from negative ones. However, such a strategy ignores the internal relationships of positive (negative) samples and often leads to overfitting, especially in the presence of hard samples and mislabeled samples. In this work, we propose a simple yet effective regularization, namely Listwise Self-Distillation (LSD), which progressively distills a model's own knowledge to adaptively assign a more appropriate distance target to each sample pair in a batch. LSD encourages smoother embeddings and information mining within positive (negative) samples as a way to mitigate overfitting and thus improve generalization. Our LSD can be directly integrated into general DML frameworks. Extensive experiments show that LSD consistently boosts the performance of various metric learning methods on multiple datasets.