Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Aug 11·5 min read
CoRCi: Cross-Reconstruction of Coherent Interests Modeling in Cross-Domain Sequential Recommendation
TL;DR
CoRCi introduces a novel framework for Cross-Domain Sequential Recommendation (CDSR) that enhances the coherence of user interests across different domains. By employing a Cross-Reconstruction approach and a new loss function, FocalNCE, CoRCi significantly outperforms existing methods in various metrics.
✦ Why It Matters
Implement CoRCi in your recommendation systems to improve user interest coherence across domains today.
Key Takeaways
- Explore CoRCi for enhancing recommendation systems across multiple domains.
- Utilize Cross-Reconstruction to improve user interest coherence.
- Implement FocalNCE to strengthen domain-invariant alignment.
- Evaluate performance improvements using real-world datasets.
How It Works
CoRCi employs a Cross-Reconstruction approach that generates mixed-domain representations from pre-encoded domain-specific data using cross-attention mechanisms. This allows the model to effectively capture and maintain coherence in user interests across different domains.
The training process utilizes a single, domain-agnostic loss function, which contrasts with traditional methods that use separate encoders and loss aggregations.