Summary Notes for Recommender Systems

Using per-item features

 

Let's add one more term to our cost function—a regularization term to help prevent overfitting. This term uses our standard regularization parameter, λ, divided by 2m(j), and multiplies that by the sum of the squares of the parameters w.

The regularization term penalizes large weight values, which in turn discourages the model from becoming overly complex. When a model has very large weights, it tends to fit the noise in the training data rather than the underlying pattern—a phenomenon known as overfitting. By adding a penalty (proportional to the squared magnitude of the weights) to the cost function, the model is encouraged to keep these weights small, leading to a simpler model that generalizes better to unseen data.

Collaborative filtering algorithm 

Collaborative filtering is effective because it leverages ratings from multiple users for the same movie, enabling us to infer potential values for its features.

Binary Labels: Favs, likes and clicks