Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

When SparseMoE Meets Noisy Interactions: An Ensemble View on Denoising Recommendation

Weipu Chen, Zhuangzhuang He, Fei Liu

TL;DR

Learning from implicit feedback is challenged by noise in user interactions. The authors propose Adaptive Ensemble Learning (AEL), which stacks sub-Autoencoders into three parent-AEs and uses a sparse gating network to adaptively select two suitable experts per input. They introduce a corrupt module and two regularizers to promote robustness and balanced expert usage. Empirical results on MovieLens, ModCloth, and Adressa demonstrate that AEL achieves state-of-the-art performance and robustness to dynamic noise, with the code released on GitHub.

Abstract

Learning user preferences from implicit feedback is one of the core challenges in recommendation. The difficulty lies in the potential noise within implicit feedback. Therefore, various denoising recommendation methods have been proposed recently. However, most of them overly rely on the hyperparameter configurations, inevitably leading to inadequacies in model adaptability and generalization performance. In this study, we propose a novel Adaptive Ensemble Learning (AEL) for denoising recommendation, which employs a sparse gating network as a brain, selecting suitable experts to synthesize appropriate denoising capacities for different data samples. To address the ensemble learning shortcoming of model complexity and ensure sub-recommender diversity, we also proposed a novel method that stacks components to create sub-recommenders instead of directly constructing them. Extensive experiments across various datasets demonstrate that AEL outperforms others in kinds of popular metrics, even in the presence of substantial and dynamic noise. Our code is available at https://github.com/cpu9xx/AEL.

When SparseMoE Meets Noisy Interactions: An Ensemble View on Denoising Recommendation

TL;DR

Learning from implicit feedback is challenged by noise in user interactions. The authors propose Adaptive Ensemble Learning (AEL), which stacks sub-Autoencoders into three parent-AEs and uses a sparse gating network to adaptively select two suitable experts per input. They introduce a corrupt module and two regularizers to promote robustness and balanced expert usage. Empirical results on MovieLens, ModCloth, and Adressa demonstrate that AEL achieves state-of-the-art performance and robustness to dynamic noise, with the code released on GitHub.

Abstract

Learning user preferences from implicit feedback is one of the core challenges in recommendation. The difficulty lies in the potential noise within implicit feedback. Therefore, various denoising recommendation methods have been proposed recently. However, most of them overly rely on the hyperparameter configurations, inevitably leading to inadequacies in model adaptability and generalization performance. In this study, we propose a novel Adaptive Ensemble Learning (AEL) for denoising recommendation, which employs a sparse gating network as a brain, selecting suitable experts to synthesize appropriate denoising capacities for different data samples. To address the ensemble learning shortcoming of model complexity and ensure sub-recommender diversity, we also proposed a novel method that stacks components to create sub-recommenders instead of directly constructing them. Extensive experiments across various datasets demonstrate that AEL outperforms others in kinds of popular metrics, even in the presence of substantial and dynamic noise. Our code is available at https://github.com/cpu9xx/AEL.
Paper Structure (11 sections, 10 equations, 4 figures, 5 tables)

This paper contains 11 sections, 10 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Proposed Model
  3. The Components: Sub-Autoencoders
  4. The Experts: Parent-Autoencoders
  5. The Brain: Sparse Gating Network
  6. Model Size discussion
  7. Experiment
  8. Experimental Settings
  9. Overall Performance Comparison
  10. Extensive Effectiveness Analysis
  11. CONCLUSION

Figures (4)

  • Figure 1: Implicit feedback brings lots of noise in the recommendation task. Denoising recommendation becomes a hotpot.
  • Figure 2: An overview of proposed AEL.
  • Figure 3: Relationship between performance and noise intensity.
  • Figure 4: Performance comparison of aggregation methods.