A Survey on Self-Attention Generative Adversarial Networks Based Tag Recommendations
Download as PDF
Chen Yuan, Xiao Zhiting, Yu Changren
Based on the analysis of previous research, this article deeply discusses the current tag-based recommendation methods from the above two aspects, briefly explains the current common tag recommendation methods and existing problems, and proposes a self-attention-based GAN (called SAGAN) tag recommendation method. In this paper, the performance index of the comparison experiment of the label recommendation approach is designed, our method is to implement self-attention mechanism to generative adversarial networks. The method of generating negative samples by the generator and identifying the samples by the discriminator enables the SAGAN to learn the pros and cons of the samples, and then use the method of generating an adversarial network to find and extract user and resource label data from the existing tags of the recommended object Characteristics, so as to better implement the recommendation that the user matches the resource tags.. Experimental results show that for different types of resources, SAGAN tag recommendation approach have achieved certain performance improvements.
Tag recommendation; sagan