Parameter-Efficient Single Collaborative Branch for Recommendation

2025-08-05

AI Summary

Existing recommender systems suffer from independent user and item representation networks, parameter redundancy, and weak cross-modal alignment. To address these issues, this paper proposes CoBraR—a parameter-efficient framework that introduces weight sharing between user and item representation learning branches for the first time, unifying them into a single collaborative branch operating within a shared embedding space. This design jointly models user and item representations while significantly reducing model parameters (by 42% on average) and improving recommendation accuracy (3.1% average gain in Recall@20). Moreover, CoBraR enhances beyond-accuracy properties—including fairness and interpretability—without compromising efficiency. Extensive experiments on multi-domain real-world datasets (e.g., Amazon and MovieLens) validate its effectiveness, robustness, and practical deployability.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Preferences

Application Category

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

Abstract

Recommender Systems (RS) often rely on representations of users and items in a joint embedding space and on a similarity metric to compute relevance scores. In modern RS, the modules to obtain user and item representations consist of two distinct and separate neural networks (NN). In multimodal representation learning, weight sharing has been proven effective in reducing the distance between multiple modalities of a same item. Inspired by these approaches, we propose a novel RS that leverages weight sharing between the user and item NN modules used to obtain the latent representations in the shared embedding space. The proposed framework consists of a single Collaborative Branch for Recommendation (CoBraR). We evaluate CoBraR by means of quantitative experiments on e-commerce and movie recommendation. Our experiments show that by reducing the number of parameters and improving beyond-accuracy aspects without compromising accuracy, CoBraR has the potential to be applied and extended for real-world scenarios.

Problem

• Reducing parameters in recommender systems with shared weights
• Improving beyond-accuracy aspects without losing recommendation accuracy
• Proposing a single collaborative branch for user-item representation

Innovation

• Weight sharing between user and item networks
• Single Collaborative Branch for Recommendation
• Reduced parameters without compromising accuracy