Abstract: Recently, purely accuracy-based recommendation algorithms no longer meet the growing diverse needs of users, because these algorithms treat all users equally, leading to the result being homogeneous. To solve this problem, we propose a gated graph walk network (Gated-GWN) consisting of two graph walk networks (GWNs) and a gated network. GWN expands a new neighborhood on the original neighborhood and aggregates the information from these neighborhoods to generate the result with accuracy or diversity. The gated network selects two results with different preferences to obtain the final results. Unlike other recommendation diversity algorithms, the accuracy-diversity ratio of the result is adjusted by the hyperparameter \lambda in GWN, rather than being entirely determined by the algorithm. With experiments on real-world datasets, we verify the effectiveness of the Gated-GWN in diversifying overall collaborative recommendations.