Matrix Factorization

Factorizing smaller matricises from a bigger Matrix.

In Recommender Systems

A Ratings-Matrix is factorized to a User and Item matrix to deal with Curse of Dimensionality. Using the Feature Importance, only the most relevant features are used to approximate the orginal matrix.

SVD and/or PCA

• Singular Value Decomposition SVD is used to split a matrix
• Principal Component Analysis PCA is a method for Dimensionality Reduction
• Both return kind of the same result

Simon Funk’s SVD

• Doesn’t replace null values