关键词: 推荐系统, 多维梯度反馈, 生成对抗网络, 协同过滤

Abstract: Recently recommender systems become more and more significant in the daily life. The combination of Generative Adversarial Networks (GAN) and recommended algorithm provides new opportunities for the development of this field. This paper identifies a unique problem about using generative adversarial networks (GAN) for collaborative filtering (CF). It does not provide comprehensive suggestion to the generator, causing problems such as an unstable performance of the generator, slow iteration speed of the model and so on, which reduces the overall recommendation effectiveness. MGFGAN was proposed to solve above problems. According to the type of generated data, the model incorporates AutoEncoder in the discriminator to provide more diversified feedback to the generator, aiming to make the generated data more closely match the user's preferences of the model. In addition, MGFGAN introduces a negative sampling module in the generator, which makes the generator take into account both items of interest and disinterest to the user, thus accelerating the generator to quickly generate data consistent with the real user distribution and improving the efficiency of the model. Finally, the experimental simulation is carried out on the public datasets FilmTrust and Ciaos based on MGFGAN. The experimental results show that the recommendation performance of MGFGAN outperforms other recommendation models based on GAN and achieves improvements in efficiency and stability.

Key words: recommendation system, multi-dimensional gradient feedback, generative adversarial networks, collaborative filter