用双曲空间提升小样本增量学习的层级表征能力。
Hyperbolic Coarse-to-Fine Few-Shot Class-Incremental Learning
- 将特征提取器嵌入双曲空间，利用庞加莱球模型实现层级数据表征。
- 在少样本条件下，粗类与细类准确率均显著提升，尤其细类增益明显。
- 适合处理具有层次结构的小样本增量学习任务，如图像分类场景。
在机器学习领域，双曲空间相较于传统欧氏空间对层级数据具有更优的表征能力。本文聚焦于粗粒度到细粒度的小样本增量学习（C2FSCIL）任务。受Knowe方法启发，该方法通过对比学习粗类标签，并在嵌入空间中归一化和冻结已学细类的分类器权重。为更好理解“粗到细”范式，我们提出将特征提取器嵌入双曲空间，采用庞加莱球模型，使特征提取器将输入图像映射至庞加莱球内的特征向量而非欧氏空间。进一步引入双曲对比损失与双曲全连接层，以支持双曲空间中的模型优化与分类。此外，为增强少样本条件下的性能，我们在双曲空间中引入最大熵分布来估计细类特征向量的概率分布，从而生成增强特征以缓解小样本训练中的过拟合问题。在C2FSCIL基准上的实验表明，该方法有效提升了粗类与细类的分类准确率。
原文摘要 · Abstract (English)
In the field of machine learning, hyperbolic space demonstrates superior representation capabilities for hierarchical data compared to conventional Euclidean space. This work focuses on the Coarse-To-Fine Few-Shot Class-Incremental Learning (C2FSCIL) task. Our study follows the Knowe approach, which contrastively learns coarse class labels and subsequently normalizes and freezes the classifier weights of learned fine classes in the embedding space. To better interpret the "coarse-to-fine" paradigm, we propose embedding the feature extractor into hyperbolic space. Specifically, we employ the Poincaré ball model of hyperbolic space, enabling the feature extractor to transform input images into feature vectors within the Poincaré ball instead of Euclidean space. We further introduce hyperbolic contrastive loss and hyperbolic fully-connected layers to facilitate model optimization and classification in hyperbolic space. Additionally, to enhance performance under few-shot conditions, we implement maximum entropy distribution in hyperbolic space to estimate the probability distribution of fine-class feature vectors. This allows generation of augmented features from the distribution to mitigate overfitting during training with limited samples. Experiments on C2FSCIL benchmarks show that our method effectively improves both coarse and fine class accuracies.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。