20240930 - RS - CF - (2)

The video discussed the use of SVD for dimensionality reduction and matrix factorization to detect latent factors and improve prediction quality.
SVD is used for dimensionality reduction in matrix factorization, selecting the best dimensions for calculation.
Matrix factorization generates low rank approximations of matrices to detect latent factors.
Prediction quality in matrix factorization can decrease due to not considering original ratings, but it can also filter out noise in the data.
The number of singular values selected in SVD depends on the desired amount of data reduction.