Recommendation of Music Teaching Content in Universities Based on Recurrent Neural Network

In order to solve the problems of homogenization and lack of personalization in university music teaching systems, a recommendation algorithm based on recurrent neural networks for university music teaching systems is proposed. The recurrent neural network model is used to estimate the implicit factor features of university music learning resources based on their textual information and to complete recommendations. In order to train the neural network model, a topic model is used as its input and an implicit factor model using L1 -norm sparse prior is proposed as the output. And a split Bregman iteration method is introduced as an optimization method in order to solve the implicit factor model for sparse prior fishing bundles. The experimental results show that the content-based methods all achieve good results, among which the proposed recurrent neural network method achieves the best results in both MAE and RMSE metrics. When the number of pooling regions is larger than the traditional average sampling (mean pooling), i.e., #pooling regions=1, the results are better. The proposed method can solve the recommendation problem of new and unpopular objects to some extent and has a high recommendation accuracy. And the introduced split Bregman iteration method can greatly improve the training efficiency of the model.