Barlow Twins for Sequential Recommendation

2025-10-30

AI Summary

To address core challenges in sequential recommendation—including interaction sparsity, popularity bias, and the difficulty of balancing accuracy and diversity—this paper proposes BT-SR, the first framework to integrate non-contrastive self-supervised learning (specifically, Barlow Twins) into sequential recommendation. BT-SR eliminates reliance on negative sampling and hand-crafted data augmentation; instead, it leverages redundancy reduction to align users’ short-term behavioral patterns while preserving distinctions in their long-term interests. A single tunable hyperparameter enables flexible control over the accuracy–diversity trade-off. Evaluated on five public benchmarks, BT-SR achieves significant improvements in next-item prediction accuracy, while simultaneously enhancing coverage of long-tail items and improving recommendation calibration—thereby effectively mitigating popularity bias.

Technology Category

• Reasoning under Uncertainty: Sequential Decision Making
• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Learning to Search

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
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems

Abstract

Sequential recommendation models must navigate sparse interaction data popularity bias and conflicting objectives like accuracy versus diversity While recent contrastive selfsupervised learning SSL methods offer improved accuracy they come with tradeoffs large batch requirements reliance on handcrafted augmentations and negative sampling that can reinforce popularity bias In this paper we introduce BT-SR a novel noncontrastive SSL framework that integrates the Barlow Twins redundancyreduction principle into a Transformerbased nextitem recommender BTSR learns embeddings that align users with similar shortterm behaviors while preserving longterm distinctionswithout requiring negative sampling or artificial perturbations This structuresensitive alignment allows BT-SR to more effectively recognize emerging user intent and mitigate the influence of noisy historical context Our experiments on five public benchmarks demonstrate that BTSR consistently improves nextitem prediction accuracy and significantly enhances longtail item coverage and recommendation calibration Crucially we show that a single hyperparameter can control the accuracydiversity tradeoff enabling practitioners to adapt recommendations to specific application needs

Problem

• Addresses sparse data and popularity bias in sequential recommendation
• Eliminates negative sampling and artificial data augmentation requirements
• Balances accuracy-diversity tradeoff through controllable hyperparameter optimization

Innovation

• Uses Barlow Twins redundancy-reduction in Transformer
• Learns embeddings without negative sampling or augmentations
• Controls accuracy-diversity tradeoff with single hyperparameter