用于推荐系统中偏好理解的因果分离变分自编码器（英文版）

英文标题：Causal Disentangled Variational Auto-Encoder for Preference Understanding in Recommendation中文摘要：本文提出了一种新方法 - 因果分离变分自编码器（CaD-VAE），该方法可以从交互数据中学习因果分离表示，以改善推荐模型的鲁棒性、可解释性和可控性，结果表明此方法可以优于现有的方法英文摘要：Recommendation models are typically trained on observational user interactiondata, but the interactions between latent factors in users' decision-makingprocesses lead to complex and entangled data. Disentangling these latentfactors to uncover their underlying representation can improve the robustness,i