基于对抗自编码器的标签推荐算法
首发时间:2019-11-26
摘要:为解决标签推荐系统面临的数据稀疏和冷启动的问题,本文提出使用对抗自编码器(Adversarial Autoencoder, AAE)探索大规模标签推荐数据集上的标签与物品之间的复杂共现关系,以进行深度协同过滤。另外,引入标签平滑以缓解过拟合,并通过修改真实概率构造来适应标签推荐系统。最后,在三个真实数据集上进行的实验表明,本文提出的方法明显优于当前已有方法。
For information in English, please click here
Adversarial autoencoder-based tag recommendation algorithm
Abstract:In order to solve the problem of data sparseness and cold start in the tag recommendation system, we proposes to use the Adversarial Autoencoder (AAE) to explore the complex co-occurrence relationship between tags and items on the large-scale tag recommendation dataset in deep collaborative filtering. In addition, label smoothing was introduced to alleviate overfitting and to adapt to the tag recommendation system by modifying the construction of true probabilities. Finally, experiments on three real datasets show that the proposed method is significantly better than the current methods.
Keywords: artificial intelligence adversarial autoencoder tag recommendation deep collaborative filtering
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于对抗自编码器的标签推荐算法
评论
全部评论0/1000