Cluster-Aware Similarity Diffusion for Instance Retrieval

2024-06-04

International Conference on Machine Learning

AI Summary

In diffusion-based re-ranking, similarity propagation is distorted by outliers and heterogeneous manifolds in the neighborhood graph. To address this, we propose a cluster-aware local similarity diffusion mechanism. Our method first constructs homogeneous local clusters to confine diffusion within structurally consistent neighborhoods. It then enforces bidirectional symmetric diffusion constraints and neighbor-guided similarity smoothing to explicitly suppress erroneous cross-manifold propagation. Finally, it models robust similarity transfer via graph neural diffusion. The core contributions are: (1) the first intra-cluster constrained diffusion paradigm, restricting similarity propagation exclusively within homogeneous local clusters; and (2) the integration of bidirectional symmetry and neighbor-guided smoothing to mitigate manifold heterogeneity. Evaluated on instance retrieval and person re-identification benchmarks, our approach achieves significant improvements in mAP and Rank-1 accuracy. The implementation is publicly available.

Technology Category

• Search and Optimization: Local Search
• Machine Learning: Learning with Manifolds
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Diffusion-based re-ranking is a common method used for retrieving instances by performing similarity propagation in a nearest neighbor graph. However, existing techniques that construct the affinity graph based on pairwise instances can lead to the propagation of misinformation from outliers and other manifolds, resulting in inaccurate results. To overcome this issue, we propose a novel Cluster-Aware Similarity (CAS) diffusion for instance retrieval. The primary concept of CAS is to conduct similarity diffusion within local clusters, which can reduce the influence from other manifolds explicitly. To obtain a symmetrical and smooth similarity matrix, our Bidirectional Similarity Diffusion strategy introduces an inverse constraint term to the optimization objective of local cluster diffusion. Additionally, we have optimized a Neighbor-guided Similarity Smoothing approach to ensure similarity consistency among the local neighbors of each instance. Evaluations in instance retrieval and object re-identification validate the effectiveness of the proposed CAS, our code is publicly available.

Problem

• Anomaly Handling
• Diffusion Methods
• Graph Similarity

Innovation

• Cluster-Aware Similarity (CAS)
• Bidirectional Method
• Object Retrieval and Re-identification