Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation

Chendi Ge, Xin Wang, Ziwei Zhang, Yijian Qin, Hong Chen, Haiyang Wu, Yang Zhang, Yuekui Yang, Wenwu Zhu

TL;DR

BiGNAS tackles cross-domain recommendation under data sparsity by jointly learning GNN architectures and behavior data importance. It introduces a Cross-Domain Customized Supernetwork for domain-specific architecture search and a Graph-Based Behavior Importance Perceptron to weigh source-domain behaviors via bi-level optimization with implicit gradients. The method demonstrates consistent improvements over state-of-the-art baselines on benchmark CDR datasets and a large-scale industry dataset, especially in sparse transfer tasks. The work advances practical cross-domain recommendations by enabling automatic, end-to-end architecture adaptation and data-driven transfer.

Abstract

Cross-domain recommendation (CDR) mitigates data sparsity and cold-start issues in recommendation systems. While recent CDR approaches using graph neural networks (GNNs) capture complex user-item interactions, they rely on manually designed architectures that are often suboptimal and labor-intensive. Additionally, extracting valuable behavioral information from source domains to improve target domain recommendations remains challenging. To address these challenges, we propose Behavior importance-aware Graph Neural Architecture Search (BiGNAS), a framework that jointly optimizes GNN architecture and data importance for CDR. BiGNAS introduces two key components: a Cross-Domain Customized Supernetwork and a Graph-Based Behavior Importance Perceptron. The supernetwork, as a one-shot, retrain-free module, automatically searches the optimal GNN architecture for each domain without the need for retraining. The perceptron uses auxiliary learning to dynamically assess the importance of source domain behaviors, thereby improving target domain recommendations. Extensive experiments on benchmark CDR datasets and a large-scale industry advertising dataset demonstrate that BiGNAS consistently outperforms state-of-the-art baselines. To the best of our knowledge, this is the first work to jointly optimize GNN architecture and behavior data importance for cross-domain recommendation.

Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation

TL;DR

BiGNAS tackles cross-domain recommendation under data sparsity by jointly learning GNN architectures and behavior data importance. It introduces a Cross-Domain Customized Supernetwork for domain-specific architecture search and a Graph-Based Behavior Importance Perceptron to weigh source-domain behaviors via bi-level optimization with implicit gradients. The method demonstrates consistent improvements over state-of-the-art baselines on benchmark CDR datasets and a large-scale industry dataset, especially in sparse transfer tasks. The work advances practical cross-domain recommendations by enabling automatic, end-to-end architecture adaptation and data-driven transfer.

Abstract

Cross-domain recommendation (CDR) mitigates data sparsity and cold-start issues in recommendation systems. While recent CDR approaches using graph neural networks (GNNs) capture complex user-item interactions, they rely on manually designed architectures that are often suboptimal and labor-intensive. Additionally, extracting valuable behavioral information from source domains to improve target domain recommendations remains challenging. To address these challenges, we propose Behavior importance-aware Graph Neural Architecture Search (BiGNAS), a framework that jointly optimizes GNN architecture and data importance for CDR. BiGNAS introduces two key components: a Cross-Domain Customized Supernetwork and a Graph-Based Behavior Importance Perceptron. The supernetwork, as a one-shot, retrain-free module, automatically searches the optimal GNN architecture for each domain without the need for retraining. The perceptron uses auxiliary learning to dynamically assess the importance of source domain behaviors, thereby improving target domain recommendations. Extensive experiments on benchmark CDR datasets and a large-scale industry advertising dataset demonstrate that BiGNAS consistently outperforms state-of-the-art baselines. To the best of our knowledge, this is the first work to jointly optimize GNN architecture and behavior data importance for cross-domain recommendation.

Paper Structure

This paper contains 20 sections, 12 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Cross-domain Recommendation
  4. GNNs and Graph Neural Architecture Search
  5. The Proposed Method
  6. Problem Formulation
  7. Cross-Domain Graph Neural Architecture Search
  8. Graph-Based Behavior Importance Perceptron
  9. Click-Through Rate Predictor
  10. Experiments
  11. Experimental Setup
  12. Datasets
  13. Baseline Models
  14. Evaluation Metric
  15. Implementation
  16. ...and 5 more sections

Figures (2)

  • Figure 1: The framework of BiGNAS is illustrated using dual-domain CDR. User behavior data from both domains is combined into a unified heterogeneous interaction graph, which serves as the model's input. The model adopts a bi-level structure: the inner Cross-Domain Customized Supernetwork follows a one-shot, retrain-free paradigm to tailor the optimal GNN architecture for each domain, while the outer Graph-Based Behavior Importance Perceptron dynamically evaluates the importance of source domain user interactions through auxiliary learning, guiding model optimization. The two modules are trained alternately to enable end-to-end optimization. The embedding $\mathbf{e}^{(i)}$ is generated by the $i$-th layer of the supernetwork.
  • Figure 2: The recommendation performance of different variants of BiGNAS.