Learning Disentangled User Representation Based on Controllable VAE for Recommendation

Yunyi Li, Pengpeng Zhao, Deqing Wang, Xuefeng Xian, Yanchi Liu, Victor S. Sheng

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

User behaviour on purchasing is always driven by complex latent factors, which are highly disentangled in the real world. Learning latent factorized representation of users can uncover user intentions behind the observed data (i.e. user-item interaction) and improve the robustness and interpretability of the recommender system. However, existing collaborative filtering methods learning disentangled representation face problems of balancing the trade-off between reconstruction quality and disentanglement. In this paper, we propose a controllable variational autoencoder framework for collaborative filtering. Specifically, we adopt a modified Proportional-Integral-Derivative (PID) control to the β -VAE objective to automatically tune the hyperparameter β using the output of Kullback-Leibler divergence as feedback. We further introduce item embeddings to guide the system to learn representation related to the real-world concepts using a factorized Gaussian distribution. Experimental results show that our model can get a crucial improvement over state-of-the-art baselines. We further evaluate our model’s effectiveness to control the trade-off between reconstruction error and disentanglement quality in the recommendation.

Original languageEnglish
Title of host publicationDatabase Systems for Advanced Applications - 26th International Conference, DASFAA 2021, Proceedings
EditorsChristian S. Jensen, Ee-Peng Lim, De-Nian Yang, Wang-Chien Lee, Vincent S. Tseng, Vana Kalogeraki, Jen-Wei Huang, Chih-Ya Shen
PublisherSpringer Science and Business Media Deutschland GmbH
Pages179-194
Number of pages16
ISBN (Print)9783030731991
DOIs
StatePublished - 2021
Event26th International Conference on Database Systems for Advanced Applications, DASFAA 2021 - Taipei, Taiwan, Province of China
Duration: Apr 11 2021Apr 14 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12683 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference26th International Conference on Database Systems for Advanced Applications, DASFAA 2021
CountryTaiwan, Province of China
CityTaipei
Period04/11/2104/14/21

Keywords

  • Disentangled representation learning
  • Recommender system
  • Variational autoencoder

Fingerprint Dive into the research topics of 'Learning Disentangled User Representation Based on Controllable VAE for Recommendation'. Together they form a unique fingerprint.

Cite this