Collaborative Filtering Recommendation Combining Adversarial Network and Attention Mechanism

Generative Adversarial Networks is an interesting way to teach computers to do human work. A strong competitor makes you better, and learning from each other in competition with him, this is the principle of GAN. In this article, we first use the user and movie information as input, and let the generator generate an n-dimensional vector. Next, we propose a new method that integrates the attention mechanism, which is to add the attention mechanism model to the adversarial network model, and let the n vector and the weight value obtained by the attention mechanism be summed and the obtained result is used as the output To make recommendations to users, we proposed the ACFGAN model. Finally, through a large number of experiments on real data sets, we verified that ACFGAN can indeed effectively improve the accuracy of recommendations.