Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cross-Domain Causal Preference Learning for Out-of-Distribution Recommendation

Zhuhang Li, Ning Yang

TL;DR

Recommender systems often fail to generalize when user attributes evolve, causing distribution shifts. The paper presents CDCOR, a cross-domain causal framework that learns domain-shared preferences via domain-adversarial training and infers a causal structure over latent user attributes to obtain invariant preferences across domains. By transferring knowledge from a rich source domain to a sparse target domain and modeling a causal process, CDCOR improves OOD recommendations and mitigates data sparsity. Experiments on Douban and Tenrec show CDCOR achieves superior OOD performance and robustness to distribution shifts, with ablations confirming the value of both cross-domain transfer and latent causal learning.

Abstract

Recommender systems use users' historical interactions to learn their preferences and deliver personalized recommendations from a vast array of candidate items. Current recommender systems primarily rely on the assumption that the training and testing datasets have identical distributions, which may not hold true in reality. In fact, the distribution shift between training and testing datasets often occurs as a result of the evolution of user attributes, which degrades the performance of the conventional recommender systems because they fail in Out-of-Distribution (OOD) generalization, particularly in situations of data sparsity. This study delves deeply into the challenge of OOD generalization and proposes a novel model called Cross-Domain Causal Preference Learning for Out-of-Distribution Recommendation (CDCOR), which involves employing a domain adversarial network to uncover users' domain-shared preferences and utilizing a causal structure learner to capture causal invariance to deal with the OOD problem. Through extensive experiments on two real-world datasets, we validate the remarkable performance of our model in handling diverse scenarios of data sparsity and out-of-distribution environments. Furthermore, our approach surpasses the benchmark models, showcasing outstanding capabilities in out-of-distribution generalization.

Cross-Domain Causal Preference Learning for Out-of-Distribution Recommendation

TL;DR

Recommender systems often fail to generalize when user attributes evolve, causing distribution shifts. The paper presents CDCOR, a cross-domain causal framework that learns domain-shared preferences via domain-adversarial training and infers a causal structure over latent user attributes to obtain invariant preferences across domains. By transferring knowledge from a rich source domain to a sparse target domain and modeling a causal process, CDCOR improves OOD recommendations and mitigates data sparsity. Experiments on Douban and Tenrec show CDCOR achieves superior OOD performance and robustness to distribution shifts, with ablations confirming the value of both cross-domain transfer and latent causal learning.

Abstract

Recommender systems use users' historical interactions to learn their preferences and deliver personalized recommendations from a vast array of candidate items. Current recommender systems primarily rely on the assumption that the training and testing datasets have identical distributions, which may not hold true in reality. In fact, the distribution shift between training and testing datasets often occurs as a result of the evolution of user attributes, which degrades the performance of the conventional recommender systems because they fail in Out-of-Distribution (OOD) generalization, particularly in situations of data sparsity. This study delves deeply into the challenge of OOD generalization and proposes a novel model called Cross-Domain Causal Preference Learning for Out-of-Distribution Recommendation (CDCOR), which involves employing a domain adversarial network to uncover users' domain-shared preferences and utilizing a causal structure learner to capture causal invariance to deal with the OOD problem. Through extensive experiments on two real-world datasets, we validate the remarkable performance of our model in handling diverse scenarios of data sparsity and out-of-distribution environments. Furthermore, our approach surpasses the benchmark models, showcasing outstanding capabilities in out-of-distribution generalization.
Paper Structure (22 sections, 17 equations, 5 figures, 4 tables)

This paper contains 22 sections, 17 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Problem Definition.
  4. Out-of-Distribution Recommendation.
  5. Proposed Method
  6. Domain-Specific Embedding
  7. Domain-Shared Embedding
  8. Causal Structure Learning
  9. Prediction and Training
  10. Experiments
  11. Experimental Setup
  12. Datasets.
  13. Baselines.
  14. Metrics and Implementation Details.
  15. Performance Comparison (RQ1 & RQ2)
  16. ...and 7 more sections

Figures (5)

  • Figure 1: Example of OOD recommendation.
  • Figure 2: The Architecture of CDCOR. The source domain is in yellow, the target domain is in blue, and the domain-shared module and causal structure learning module is in green.
  • Figure 3: Robustness experiments. (a) OOD#1 test in Douban. (b) OOD#1 test in Tenrec. (c) OOD#2 test in Tenrec.
  • Figure 4: Performance against sparsity in Tenrec. (a) IID test. (b) OOD#1 test. (c) OOD#2 test.
  • Figure 5: Performance against sparsity in Douban. (a) IID test. (b) OOD#1 test