Learning from implicit feedback is one of the most common cases in the application of recommender systems. Generally speaking, interacted examples are considered as positive while negative examples are sampled from uninteracted ones. However, noisy examples are prevalent in real-world implicit feedback. A noisy positive example could be interacted but it actually leads to negative user preference. A noisy negative example which is uninteracted because of unawareness of the user could also denote potential positive user preference. Conventional training methods overlook these noisy examples, leading to sub-optimal recommendation. In this work, we propose probabilistic and variational recommendation denoising for implicit feedback. Through an empirical study, we find that different models make relatively similar predictions on clean examples which denote the real user preference, while the predictions on noisy examples vary much more across different models. Motivated by this observation, we propose denoising with probabilistic inference (DPI) which aims to minimize the KL-divergence between the real user preference distributions parameterized by two recommendation models while maximize the likelihood of data observation. We then show that DPI recovers the evidence lower bound of an variational auto-encoder when the real user preference is considered as the latent variables. This leads to our second learning framework denoising with variational autoencoder (DVAE). We employ the proposed DPI and DVAE on four state-of-the-art recommendation models and conduct experiments on three datasets. Experimental results demonstrate that DPI and DVAE significantly improve recommendation performance compared with normal training and other denoising methods. Codes will be open-sourced.
翻译:从隐性反馈中学习的隐性反馈是应用推荐者系统中最常见的案例之一。 一般来说, 互动实例被视为是积极的, 而负面实例则取自非互动的样本。 然而, 噪音实例在现实世界的隐含反馈中很普遍。 一个吵闹的正面实例可以互动, 但实际上导致用户偏好。 一个因用户不知道而未互动的吵闹负面实例也可能表示潜在的积极用户偏好。 常规培训方法忽略了这些吵闹的例子, 导致次优化建议。 在这项工作中, 我们建议对隐含反馈进行概率和变异性建议, 并提议对隐性建议进行分解。 我们通过经验研究发现, 不同的模型对表明真正用户偏爱的清洁实例作出相对相似的预测, 而对杂乱实例的预测则在不同模式中差异更大。 我们提出, 以预测为目的, 将真正的公开用户偏好分配模式之间的 KL- 调调调调, 并用两个建议模型对用户偏重度进行比重度, 同时将数据观测的可能性最大化。 我们随后认为, DPI 会恢复真实的三度, 将数据偏差值比重,, 以亚值变变变 。