Retrievable Domain-Sensitive Feature Memory for Multi-Domain Recommendation

2024-05-21

arXiv.org

AI Summary

Existing multi-domain advertising recommendation methods neglect domain sensitivity in feature distribution during implicit modeling, leading to inadequate characterization of domain discrepancies. To address this, we propose a domain-sensitive feature attribution method that systematically identifies critical features exhibiting significant cross-domain distributional and predictive-effect disparities—marking the first such systematic identification. We further design a retrievable and updatable domain-sensitive feature memory module to explicitly model both domain-shared and domain-specific characteristics. Additionally, we establish a unified multi-domain joint training framework coupled with online/offline collaborative evaluation. Extensive experiments across multiple industrial advertising scenarios demonstrate an average 0.8% improvement in CTR prediction AUC and a 12% increase in online impression diversity, substantially enhancing the model’s perceptiveness and discriminative capability regarding domain differences.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Security
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

With the increase in the business scale and number of domains in online advertising, multi-domain ad recommendation has become a mainstream solution in the industry. The core of multi-domain recommendation is effectively modeling the commonalities and distinctions among domains. Existing works are dedicated to designing model architectures for implicit multi-domain modeling while overlooking an in-depth investigation from a more fundamental perspective of feature distributions. This paper focuses on features with significant differences across various domains in both distributions and effects on model predictions. We refer to these features as domain-sensitive features, which serve as carriers of domain distinctions and are crucial for multi-domain modeling. Experiments demonstrate that existing multi-domain modeling methods may neglect domain-sensitive features, indicating insufficient learning of domain distinctions. To avoid this neglect, we propose a domain-sensitive feature attribution method to identify features that best reflect domain distinctions from the feature set. Further, we design a memory architecture that extracts domain-specific information from domain-sensitive features for the model to retrieve and integrate, thereby enhancing the awareness of domain distinctions. Extensive offline and online experiments demonstrate the superiority of our method in capturing domain distinctions and improving multi-domain recommendation performance.

Problem

• Identifying domain-sensitive features in multi-domain recommendation systems
• Enhancing domain distinction awareness through feature attribution
• Improving recommendation performance with retrievable memory architecture

Innovation

• Identifies domain-sensitive features via attribution method
• Designs memory architecture for domain-specific information
• Enhances domain distinction awareness in recommendations