S2DNMF: A Self-supervised Deep Nonnegative Matrix Factorization Recommendation Model Incorporating Deep Latent Features of Network Structure

This paper presents S2DNMF, a self-supervised deep nonnegative matrix factorization recommendation model. By incorporating deep latent features derived from network structure, S2DNMF improves the recommendation accuracy and robustness in real-world applications. The model effectively integrates both collaborative filtering and network-based information, providing a more comprehensive recommendation approach.