Binary rating vs. actual rating #67
Description
In this model, all the ratings of positive samples have been normalized to 1 (instead of regular 1-5 rating score), is there any particular reason for this treatment? Would the information on user's preference on certain items (movies for instance) be lost in the model training? Thanks!