Collaborative Filtering

Collaborative filtering (CF) is a core technique in recommender systems aiming to predict user preferences by leveraging past user-item interactions. Current research emphasizes improving CF's accuracy and efficiency, particularly addressing challenges like cold-start problems (limited user data) and biases (favoring popular items or user groups), through advancements in model architectures such as graph neural networks, diffusion models, and autoencoders, as well as refined loss functions and sampling strategies. These improvements are crucial for enhancing the personalization and fairness of recommendation systems across diverse applications, from e-commerce and social media to specialized domains like course selection and real estate.