1 paper · 1 filter
Xufeng Liang, Zhida Qin, Chong Zhang +2
Contrastive learning has demonstrated promising potential in recommender systems. Existing methods typically construct sparser views by randomly perturbing the original interaction…