Exploring Transferable Homogeneous Groups for Compositional Zero-Shot Learning

2025-01-18

AI Summary

To address feature drift caused by conditional dependencies in compositional zero-shot learning, this paper proposes Homogeneous Group Representation Learning (HGRL). HGRL introduces a novel analogy-based hierarchical homogeneous grouping paradigm that adaptively aggregates semantically similar classes and jointly learns subgroup representations sharing common attributes. It employs a three-module collaborative architecture: visual feature disentanglement, prompt-aligned optimization, and intra-group consistency constraints—enabling dual-path (vision–text) representation enhancement. The learned distributed group centroids retain strong discriminability while significantly improving semantic transferability. Extensive experiments on three standard benchmarks demonstrate that HGRL consistently outperforms state-of-the-art methods, achieving substantial gains in generalization to unseen compositions. This validates that homogeneous grouping achieves a superior trade-off between transferability and discriminability.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Representation Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Conditional dependency present one of the trickiest problems in Compositional Zero-Shot Learning, leading to significant property variations of the same state (object) across different objects (states). To address this problem, existing approaches often adopt either all-to-one or one-to-one representation paradigms. However, these extremes create an imbalance in the seesaw between transferability and discriminability, favoring one at the expense of the other. Comparatively, humans are adept at analogizing and reasoning in a hierarchical clustering manner, intuitively grouping categories with similar properties to form cohesive concepts. Motivated by this, we propose Homogeneous Group Representation Learning (HGRL), a new perspective formulates state (object) representation learning as multiple homogeneous sub-group representation learning. HGRL seeks to achieve a balance between semantic transferability and discriminability by adaptively discovering and aggregating categories with shared properties, learning distributed group centers that retain group-specific discriminative features. Our method integrates three core components designed to simultaneously enhance both the visual and prompt representation capabilities of the model. Extensive experiments on three benchmark datasets validate the effectiveness of our method.

Problem

• Combining Zero-Shot Learning
• Conditional Dependence
• Feature Variation

Innovation

• Homogeneous Group Representation Learning
• Visual Recognition
• Text Understanding