Feature Identification for Hierarchical Contrastive Learning
This addresses the problem of neglecting hierarchical relationships in classification for computer vision applications, representing an incremental advance over existing hierarchical contrastive learning methods.
The paper tackles hierarchical classification by proposing two hierarchical contrastive learning methods that model inter-class relationships and imbalanced distributions, achieving state-of-the-art performance with a 2 percentage point accuracy improvement on CIFAR100 and ModelNet40 datasets.
Hierarchical classification is a crucial task in many applications, where objects are organized into multiple levels of categories. However, conventional classification approaches often neglect inherent inter-class relationships at different hierarchy levels, thus missing important supervisory signals. Thus, we propose two novel hierarchical contrastive learning (HMLC) methods. The first, leverages a Gaussian Mixture Model (G-HMLC) and the second uses an attention mechanism to capture hierarchy-specific features (A-HMLC), imitating human processing. Our approach explicitly models inter-class relationships and imbalanced class distribution at higher hierarchy levels, enabling fine-grained clustering across all hierarchy levels. On the competitive CIFAR100 and ModelNet40 datasets, our method achieves state-of-the-art performance in linear evaluation, outperforming existing hierarchical contrastive learning methods by 2 percentage points in terms of accuracy. The effectiveness of our approach is backed by both quantitative and qualitative results, highlighting its potential for applications in computer vision and beyond.