recommender system problem formulation
—a content based recommendations
- Notations
- Algorithm(with linear regression)
- featuring learning
—Collaborative Filtering
(Low rank matrix fractorization)
iterating and optimizing both of θ and X
to solve simultaneously for x and theta
(without
x0=θ0=1
)
Vectorization
prediction=X ΘT
—Mean Normalization
if there is a person who has never rated for any movie,because of the normalization, we will get a θ with all zero.
In case of this,
then, instead of zero, the value will be replaced of a mean value which means no attitude to it.