ASPIRE: Make Spectral Graph Collaborative Filtering Great Again via Adaptive Filter Learning

2026-04-24

AI Summary

This work addresses the “low-frequency explosion” phenomenon in traditional spectral graph collaborative filtering, which stems from bias in the objective function and impedes effective graph filter learning. To resolve this issue, the authors propose the ASPIRE framework, which for the first time elucidates the fundamental impact of this problem on filter learning and introduces a theoretically grounded decoupled learning mechanism. ASPIRE employs a bilevel optimization objective to enable end-to-end adaptive learning of graph filters without requiring manual hyperparameter tuning. Extensive experiments across multiple recommendation scenarios demonstrate that the proposed method significantly enhances performance, spectral adaptability, and training stability, with the learned filters matching or even surpassing those of carefully handcrafted baselines.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Graph filter design is central to spectral collaborative filtering, yet most existing methods rely on manually tuned hyperparameters rather than fully learnable filters. We show that this challenge stems from a bias in traditional recommendation objectives, which induces a spectral phenomenon termed low-frequency explosion, thereby fundamentally hindering the effective learning of graph filters. To overcome this limitation, we propose a novel adaptive spectral graph collaborative filtering framework (ASPIRE) based on a bi-level optimization objective. Guided by our theoretical analysis, we disentangle the filter learning objective, which in turn leads to excellent recommendation performance, spectral adaptivity, and training stability in practice. Extensive experiments show our learned filters match the performance of carefully engineered task-specific designs. Furthermore, ASPIRE is equally effective in LLM-powered collaborative filtering. Our findings demonstrate that graph filter learning is viable and generalizable, paving the way for more expressive graph neural networks in collaborative filtering.

Problem

• spectral collaborative filtering
• graph filter learning
• low-frequency explosion
• recommendation systems
• adaptive filtering

Innovation

• adaptive filter learning
• spectral graph collaborative filtering
• low-frequency explosion
• bi-level optimization
• graph neural networks