Recommender Systems: Overview and Methods

This lecture covers the motivation behind the prevalence of recommenders today, the difference between collaborative filtering and content-based recommenders, the Netflix Prize competition, neighborhood methods, latent factor methods, overfitting, regularization, and stochastic gradient descent. It discusses the evolution from traditional to online retailers, the importance of filtering and choosing in online services, and the role of recommenders in platforms like Amazon, YouTube, and Spotify. The lecture also delves into collaborative filtering, baseline predictors, model training, regularization techniques, model validation, and the concepts of neighborhood models, similarity metrics, and latent factor models. It concludes with a comparison of stochastic gradient descent and alternating least squares for optimizing latent factor models.