Learning Disentangled Representations for Recommendation

User behavior data in recommender systems are driven by the complex interactions of many latent factors behind the users' decision making processes.