Negative Margin Matters: Understanding Margin in Few-shot Classification

By Bin Liu, Yue Cao, Yutong Lin, Qi Li, Zheng Zhang, Mingsheng Long, Han Hu. 2020

380 citations recorded by the source corpus.

This paper introduces a negative margin loss to metric learning based few-shot learning methods. The negative margin loss significantly outperforms regular softmax loss, and achieves state-of-the-art accuracy on three standard few-shot classification benchmarks with few bells and whistles. These results are contrary to the common practice in the metric learning field, that the margin is zero or positive. To understand why the negative margin loss performs well for the few-shot classification, we analyze the discriminability of learned features w.r.t different margins for training and novel classes, both empirically and theoretically. We find that although negative margin reduces the feature discriminability for training classes, it may also avoid falsely mapping samples of the same novel class to multiple peaks or clusters, and thus benefit the discrimination of novel classes. Code is available at this https URL.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Margin-Based Few-Shot Class-Incremental Learning with Class-Level Overfitting Mitigation (2022, methodology): “For such dilemma, [24] gave an explanation by the degraded mapping from novel to base classes.”
Forward Compatible Few-Shot Class-Incremental Learning (2022, methodology): “• Decoupled-NegCosine [4]: Similar to DecoupledCosine, it decouples the training process of embedding and classifier.”
Why Do Better Loss Functions Lead to Less Transferable Features? (2020, background): “[37] find that introducing a negative margin into a cosine softmax loss yields both lower class separation and higher accuracy.”
A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, background): “Prototype Network [123] Prototype representation European Original Negative Margin Matters [129] Negative margin loss Cosine Balancing discriminative and migratory”
Anti-aliasing Semantic Reconstruction for Few-Shot Semantic Segmentation (2021, background): “Extra margin constraints [20, 17] are absorbed into metric learning to further adjust the inter-class diversity and intra-class variance.”
DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning (2020, background): “The research literature on few-shot learning shows great diversity [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25].”
Orthogonal Projection Loss (2021, neutral): “Note that [40, 28, 31] are recent loss functions speciﬁc to FSL. Dataset CNAPs [42] (NeurIPS’19) SUR [11] (ECCV’20”