DeNUS Outline ational University Introduction motivation Method Recap bpr Bayesian Personalized Ranking APR: Adversarial training for BPR Experiments Conclusion
Outline • Introduction & Motivation • Method – Recap BPR (Bayesian Personalized Ranking) – APR: Adversarial Training for BPR • Experiments • Conclusion 6
DeNUS Recap BPR ational University BPR aims to maximize the margin between an ordered example paIr. sigmoid Positive prediction Negative prediction LBPR(O⊙) In oyu(o)-yuj(e)+hellOll (,i,j)∈分 Pairwise training examples: u prefers i over j An example of using bpr to optimize mf model BPR Objective In al.u- 9u). Training Minimizer yui=Puq redictions Puqi Embeddin qp q (positive item) user (negative item [Rendle et al, UAI09
Recap BPR 7 • BPR aims to maximize the margin between an ordered example pair. • An example of using BPR to optimize MF model: Pairwise training examples: u prefers i over j sigmoid Positive prediction Negative prediction [Rendle et al, UAI’09]