Epistemic Uncertainty-aware Recommendation Systems via Bayesian Deep Ensemble Learning

2025-04-14

AI Summary

To address overfitting and the lack of epistemic uncertainty modeling in recommender systems under sparse explicit feedback scenarios, this paper proposes the Bayesian Deep Ensemble Collaborative Filtering (BDECF) framework. BDECF innovatively integrates Bayesian neural networks (to model weight-level epistemic uncertainty), attention-driven nonlinear embedding matching (to enhance interpretability), and an ensemble hypermodel (to improve robustness), thereby enabling epistemic uncertainty-aware recommendation. Uncertainty calibration and ensemble-based prediction jointly enhance predictive reliability and generalization. Extensive experiments on multiple real-world sparse datasets demonstrate significant improvements in Recall@K and NDCG over state-of-the-art baselines. Ablation studies validate the effectiveness of each core component. This work establishes a novel paradigm for uncertainty-aware recommendation, advancing both theoretical understanding and practical deployment in sparse-feedback settings.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• 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

Recommending items to users has long been a fundamental task, and studies have tried to improve it ever since. Most well-known models commonly employ representation learning to map users and items into a unified embedding space for matching assessment. These approaches have primary limitations, especially when dealing with explicit feedback and sparse data contexts. Two primary limitations are their proneness to overfitting and failure to incorporate epistemic uncertainty in predictions. To address these problems, we propose a novel Bayesian Deep Ensemble Collaborative Filtering method named BDECF. To improve model generalization and quality, we utilize Bayesian Neural Networks, which incorporate uncertainty within their weight parameters. In addition, we introduce a new interpretable non-linear matching approach for the user and item embeddings, leveraging the advantages of the attention mechanism. Furthermore, we endorse the implementation of an ensemble-based supermodel to generate more robust and reliable predictions, resulting in a more complete model. Empirical evaluation through extensive experiments and ablation studies across a range of publicly accessible real-world datasets with differing sparsity characteristics confirms our proposed method's effectiveness and the importance of its components.

Problem

• Address overfitting in recommendation systems with sparse data
• Incorporate epistemic uncertainty in recommendation predictions
• Improve model generalization using Bayesian Neural Networks

Innovation

• Bayesian Neural Networks for uncertainty incorporation
• Interpretable non-linear matching with attention
• Ensemble-based supermodel for robust predictions