推荐系统中矩阵分解方法：svd，非对称svd和svd++的区别

Some SVD-inspired methods used in the Netflix Prize include:

• Standard SVD: Once you've represented users and movies as factor vectors, you can dot product Alice's vector with Inception's vector to get Alice's predicted rating of Inception.
• Asymmetric SVD model: Instead of users having their own notion of factor vectors, we can represent users as a bag of items they have rated (or provided implicit feedback for). So Alice is now represented as a (possibly weighted) sum of the factor vectors of the items she has rated, and to get her predicted rating of Titanic, we can dot product this  representation with the factor vector of Titanic. From a practical perspective, this model has an added benefit in that no userparameterizations are needed, so we can use this approach to generate recommendations as soon as a user provides some feedback (which could  just be views of an item, and not necessarily ratings), without needing to retrain the model to factorize the user.
• SVD++ model: Incorporate both the standard SVD and the asymmetric SVD model by representing users both with their own factor representation and as a bag of item vectors.