Matrix Factorization Model, consists item model and user model.
Item model is a [#item, rank] matrix sotred on PS, each row represents an item.
User model a UserVecs storage, each UserVec represents a user by a rank-dim vector.
To save storage, we only cached items that appears in the local worker trainning data.
Matrix Factorization Model, consists item model and user model. Item model is a [#item, rank] matrix sotred on PS, each row represents an item. User model a UserVecs storage, each UserVec represents a user by a rank-dim vector. To save storage, we only cached items that appears in the local worker trainning data.