mlpack
blog
|
Alternatives to Neighborhood-Based CF - Week 7
For the past week, I was working on refactoring decomposition policies. Now I have finished refactoring all current decomposition policies, and made necessary modifications in CFType class, cf main program, interpolation policies, and cf tests. The major changes are: 1) model parameters (e.g. W
, H
) are moved into decomposition class, 2) decomposition class provides method GetRating()
to compute prediction and GetNeighborhood()
to compute neighborhood and similarities, 3) class CFModel
(similar to AdaBoostModel
) is added and is used for cf main program. After refactoring decomposition policies, it would be easier to implement BiasSVDPolicy
and SVDPlusPlusPolicy
(and perhaps other policies later). I will focus on implementing these two policies next week.
Generated by 1.8.13