Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unleashing the Potential of Multi-Channel Fusion in Retrieval for Personalized Recommendations

Junjie Huang, Jiarui Qin, Jianghao Lin, Ziming Feng, Yong Yu, Weinan Zhang

TL;DR

This paper is the first to identify and systematically investigate multi-channel fusion in the retrieval stage, and utilizes black-box optimization techniques, including the Cross Entropy Method and Bayesian Optimization for global weight optimization, alongside policy gradient-based approaches for personalized merging.

Abstract

Recommender systems (RS) are pivotal in managing information overload in modern digital services. A key challenge in RS is efficiently processing vast item pools to deliver highly personalized recommendations under strict latency constraints. Multi-stage cascade ranking addresses this by employing computationally efficient retrieval methods to cover diverse user interests, followed by more precise ranking models to refine the results. In the retrieval stage, multi-channel retrieval is often used to generate distinct item subsets from different candidate generators, leveraging the complementary strengths of these methods to maximize coverage. However, forwarding all retrieved items overwhelms downstream rankers, necessitating truncation. Despite advancements in individual retrieval methods, multi-channel fusion, the process of efficiently merging multi-channel retrieval results, remains underexplored. We are the first to identify and systematically investigate multi-channel fusion in the retrieval stage. Current industry practices often rely on heuristic approaches and manual designs, which often lead to suboptimal performance. Moreover, traditional gradient-based methods like SGD are unsuitable for this task due to the non-differentiable nature of the selection process. In this paper, we explore advanced channel fusion strategies by assigning systematically optimized weights to each channel. We utilize black-box optimization techniques, including the Cross Entropy Method and Bayesian Optimization for global weight optimization, alongside policy gradient-based approaches for personalized merging. Our methods enhance both personalization and flexibility, achieving significant performance improvements across multiple datasets and yielding substantial gains in real-world deployments, offering a scalable solution for optimizing multi-channel fusion in retrieval.

Unleashing the Potential of Multi-Channel Fusion in Retrieval for Personalized Recommendations

TL;DR

This paper is the first to identify and systematically investigate multi-channel fusion in the retrieval stage, and utilizes black-box optimization techniques, including the Cross Entropy Method and Bayesian Optimization for global weight optimization, alongside policy gradient-based approaches for personalized merging.

Abstract

Recommender systems (RS) are pivotal in managing information overload in modern digital services. A key challenge in RS is efficiently processing vast item pools to deliver highly personalized recommendations under strict latency constraints. Multi-stage cascade ranking addresses this by employing computationally efficient retrieval methods to cover diverse user interests, followed by more precise ranking models to refine the results. In the retrieval stage, multi-channel retrieval is often used to generate distinct item subsets from different candidate generators, leveraging the complementary strengths of these methods to maximize coverage. However, forwarding all retrieved items overwhelms downstream rankers, necessitating truncation. Despite advancements in individual retrieval methods, multi-channel fusion, the process of efficiently merging multi-channel retrieval results, remains underexplored. We are the first to identify and systematically investigate multi-channel fusion in the retrieval stage. Current industry practices often rely on heuristic approaches and manual designs, which often lead to suboptimal performance. Moreover, traditional gradient-based methods like SGD are unsuitable for this task due to the non-differentiable nature of the selection process. In this paper, we explore advanced channel fusion strategies by assigning systematically optimized weights to each channel. We utilize black-box optimization techniques, including the Cross Entropy Method and Bayesian Optimization for global weight optimization, alongside policy gradient-based approaches for personalized merging. Our methods enhance both personalization and flexibility, achieving significant performance improvements across multiple datasets and yielding substantial gains in real-world deployments, offering a scalable solution for optimizing multi-channel fusion in retrieval.

Paper Structure

This paper contains 36 sections, 30 equations, 8 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. PRELIMINARIES
  4. Problem Formulation
  5. Rationale Behind Weight Assignment
  6. Methodology
  7. Globally Unified Weight Assignment
  8. Cross Entropy Method
  9. Bayesian Optimization
  10. Personalized Weight Assignment
  11. Model Architecture
  12. Objective Function
  13. Experiments
  14. Dataset and Experimental Flow
  15. Baselines and Evaluation Metrics
  16. ...and 21 more sections

Figures (8)

  • Figure 1: Up: Illustration of multi-stage cascade ranking and multi-channel retrieval in recommender systems. Bottom: Performance variations with different weight combinations on Gowalla (left) and Amazon_Books (right).
  • Figure 2: Diversity among various candidate generators from both item and user perspectives on Amazon_Books.
  • Figure 3: An illustration of our non-personalized and personalized multi-channel fusion strategies in the retrieval stage.
  • Figure 4: Optimal weights for various retrieval channels generated by Bayesian Optimization on Amazon Books and Tmall. Proportions below 2% are omitted for clarity.
  • Figure 5: Effect of $\xi$ on Gowalla and Amazon_Books.
  • ...and 3 more figures