Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Federated Semantic Learning for Privacy-preserving Cross-domain Recommendation

Ziang Lu, Lei Guo, Xu Yu, Zhiyong Cheng, Xiaohui Han, Lei Zhu

TL;DR

This work introduces federated semantic learning and devise FFMSR, a Fast Fourier Transform-based filter and a gating mechanism are developed to alleviate the impact of irrelevant semantic information in the local model.

Abstract

In the evolving landscape of recommender systems, the challenge of effectively conducting privacy-preserving Cross-Domain Recommendation (CDR), especially under strict non-overlapping constraints, has emerged as a key focus. Despite extensive research has made significant progress, several limitations still exist: 1) Previous semantic-based methods fail to deeply exploit rich textual information, since they quantize the text into codes, losing its original rich semantics. 2) The current solution solely relies on the text-modality, while the synergistic effects with the ID-modality are ignored. 3) Existing studies do not consider the impact of irrelevant semantic features, leading to inaccurate semantic representation. To address these challenges, we introduce federated semantic learning and devise FFMSR as our solution. For Limitation 1, we locally learn items'semantic encodings from their original texts by a multi-layer semantic encoder, and then cluster them on the server to facilitate the transfer of semantic knowledge between domains. To tackle Limitation 2, we integrate both ID and Text modalities on the clients, and utilize them to learn different aspects of items. To handle Limitation 3, a Fast Fourier Transform (FFT)-based filter and a gating mechanism are developed to alleviate the impact of irrelevant semantic information in the local model. We conduct extensive experiments on two real-world datasets, and the results demonstrate the superiority of our FFMSR method over other SOTA methods. Our source codes are publicly available at: https://github.com/Sapphire-star/FFMSR.

Federated Semantic Learning for Privacy-preserving Cross-domain Recommendation

TL;DR

This work introduces federated semantic learning and devise FFMSR, a Fast Fourier Transform-based filter and a gating mechanism are developed to alleviate the impact of irrelevant semantic information in the local model.

Abstract

In the evolving landscape of recommender systems, the challenge of effectively conducting privacy-preserving Cross-Domain Recommendation (CDR), especially under strict non-overlapping constraints, has emerged as a key focus. Despite extensive research has made significant progress, several limitations still exist: 1) Previous semantic-based methods fail to deeply exploit rich textual information, since they quantize the text into codes, losing its original rich semantics. 2) The current solution solely relies on the text-modality, while the synergistic effects with the ID-modality are ignored. 3) Existing studies do not consider the impact of irrelevant semantic features, leading to inaccurate semantic representation. To address these challenges, we introduce federated semantic learning and devise FFMSR as our solution. For Limitation 1, we locally learn items'semantic encodings from their original texts by a multi-layer semantic encoder, and then cluster them on the server to facilitate the transfer of semantic knowledge between domains. To tackle Limitation 2, we integrate both ID and Text modalities on the clients, and utilize them to learn different aspects of items. To handle Limitation 3, a Fast Fourier Transform (FFT)-based filter and a gating mechanism are developed to alleviate the impact of irrelevant semantic information in the local model. We conduct extensive experiments on two real-world datasets, and the results demonstrate the superiority of our FFMSR method over other SOTA methods. Our source codes are publicly available at: https://github.com/Sapphire-star/FFMSR.

Paper Structure

This paper contains 37 sections, 25 equations, 9 figures, 8 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Privacy-preserving Cross-domain Recommendation
  4. Semantic-enhanced Recommendation
  5. Denoising in Recommendation
  6. Differences
  7. Methodology
  8. Preliminaries
  9. Overview of FFMSR
  10. Deep Semantic Extraction
  11. Semantic Encoding by a Multi-layer PLM
  12. MoE Adapter.
  13. Fusion Block
  14. Cross-domain Semantic Fusion
  15. Semantic Clustering
  16. ...and 22 more sections

Figures (9)

  • Figure 1: An example of capturing common semantic information between different items through clustering.
  • Figure 2: The system architecture of FFMSR, which is mainly composed of semantic extraction, cross-domain semantic fusion, and semantic filtering components. (1) The semantic extraction module aims at learning the semantics of items' description texts at multi-levels of PLM. The semantic encodings within it will be uploaded to the server for further fusion. (2) The cross-domain semantic fusion targets fusing the semantics in different domains by clustering. The clustered encoding will be sent back to the client for local updating. (3) The semantic Filtering intends to filter out the irrelevant semantics caused by semantic fusion and multi-modality utilization.
  • Figure 3: The structure of the fusion block with Id embeddings as supervision signals.
  • Figure 4: The improved sequence encoder with FFT-based filter layer.
  • Figure 5: Impact of the number of cluster centers $K$ on OnlineRetail-Pantry.
  • ...and 4 more figures