mlpack
blog
|
Alternatives to Neighborhood-Based CF - Week 2
For the second week, I first read some materials on computing interpolation weights for neighborhood-based collaborative filtering. There are three methods I planned to implment: average interpolation, similarity-based interpolation, and regression-based interpolation. So far I have implmented AverageInterpolation
and SimilarityInterpolation
, and I am currently working on RegressionInterpolation
. As for rating normalization, I merged codes of refactored CF class (thanks to Haritha for the refactoring work), and I am working to solve some issues that occur after the merge.
For the third week, I am going to modify methods in CF to use policy-design for weights interpolation and neighbor search, add RegressionInterpolation
. After that I will do the remaining work of rating normalization.
Generated by 1.8.13