Abstract:Cross-domain recommendation (CDR) methods are proposed to tackle the sparsity problem in click through rate (CTR) estimation. Existing CDR methods directly transfer knowledge from the source domains to the target domain and ignore the heterogeneities among domains, including feature dimensional heterogeneity and latent space heterogeneity, which may lead to negative transfer. Besides, most of the existing methods are based on single-source transfer, which cannot simultaneously utilize knowledge from multiple source domains to further improve the model performance in the target domain. In this paper, we propose a centralized-distributed transfer model (CDTM) for CDR based on multi-source heterogeneous transfer learning. To address the issue of feature dimension heterogeneity, we build a dual embedding structure: domain specific embedding (DSE) and global shared embedding (GSE) to model the feature representation in the single domain and the commonalities in the global space,separately. To solve the latent space heterogeneity, the transfer matrix and attention mechanism are used to map and combine DSE and GSE adaptively. Extensive offline and online experiments demonstrate the effectiveness of our model.

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are the data sparsity and heterogeneity problems in Cross - Domain Recommendation (CDR). Specifically: 1. **Data Sparsity Problem**: In recommendation systems and online advertising systems, many domains face performance challenges due to insufficient data. Traditional Click - Through Rate (CTR) models require a large amount of data within a single domain to train high - performance models, but in some domains with scarce data, this requirement is difficult to meet. Therefore, the idea of transfer learning is introduced, and the data of other domains are utilized through cross - domain recommendation methods to improve the recommendation effect in the target domain. 2. **Domain Heterogeneity Problem**: - **Feature - Dimension Heterogeneity**: Different domains may have different feature dimensions, that is, some features exist in some domains but not in others. - **Latent - Space Heterogeneity**: Even the same features may have different distributions in different domains. This causes the model networks of different domains not to be directly shared, thus limiting the application of existing cross - domain recommendation methods and may lead to negative transfer. To solve the above problems, the authors propose a Centralized - Distributed Transfer Model (CDTM) based on multi - source heterogeneous transfer learning. The main contributions of this model include: 1. **Centralized - Distributed Transfer Model**: This model can be extended to scenarios of more domains and simultaneously improve the performance of models in multiple domains. 2. **Dual - Embedding Structure**: Domain Specific Embedding (DSE) and Global Shared Embedding (GSE) are constructed, which are respectively used to model the unique feature representations of a single domain and the global feature representations of all domains. A combined attention mechanism is used to adaptively combine the dual - embeddings of transferable features. 3. **Transfer Matrix**: The transfer matrix is used to map GSE to the latent space shared with DSE to solve the heterogeneity problem in cross - domain recommendation. And an auxiliary loss function is used to help optimize the transfer matrix. 4. **Extensive Experimental Verification**: A large number of offline and online experiments are carried out based on real - world commercial data, which prove the effectiveness and robustness of the model. Through these innovations, the CDTM model can achieve better performance in multi - source heterogeneous cross - domain recommendation tasks.

A Centralized-Distributed Transfer Model for Cross-Domain Recommendation Based on Multi-Source Heterogeneous Transfer Learning

User Distribution Mapping Modelling with Collaborative Filtering for Cross Domain Recommendation

Transferring Causal Mechanism over Meta-representations for Target-unknown Cross-domain Recommendation

Heterogeneous Graph-based Framework with Disentangled Representations Learning for Multi-target Cross Domain Recommendation

Heterogeneous and Clustering-Enhanced Personalized Preference Transfer for Cross-Domain Recommendation

DTCDR: A Framework for Dual-Target Cross-Domain Recommendation

Combining Graph Contrastive Embedding and Multi-head Cross-Attention Transfer for Cross-Domain Recommendation

Mitigating Negative Transfer in Cross-Domain Recommendation Via Knowledge Transferability Enhancement

Towards Universal Cross-Domain Recommendation

DDTCDR: Deep Dual Transfer Cross Domain Recommendation

A Collaborative Transfer Learning Framework for Cross-domain Recommendation

AMT-CDR: A Deep Adversarial Multi-channel Transfer Network for Cross-domain Recommendation

Transfer Meets Hybrid: A Synthetic Approach for Cross-Domain Collaborative Filtering with Text

CDR-Adapter: Learning Adapters to Dig Out More Transferring Ability for Cross-Domain Recommendation Models

The Devil is in the Sources! Knowledge Enhanced Cross-Domain Recommendation in an Information Bottleneck Perspective

Learning Accurate and Bidirectional Transformation Via Dynamic Embedding Transportation for Cross-Domain Recommendation

Domain-Aware Cross-Attention for Cross-domain Recommendation

Beyond the Overlapping Users: Cross-Domain Recommendation via Adaptive Anchor Link Learning

Triple Sequence Learning for Cross-domain Recommendation

REMIT: Reinforced Multi-Interest Transfer for Cross-Domain Recommendation

DDGHM: Dual Dynamic Graph with Hybrid Metric Training for Cross-Domain Sequential Recommendation