Wide & Deep Learning for Recommender Systems

24 Jun 2016Heng-Tze ChengLevent KocJeremiah HarmsenTal ShakedTushar ChandraHrishi AradhyeGlen AndersonGreg CorradoWei ChaiMustafa IspirRohan AnilZakaria HaqueLichan HongVihan JainXiaobing LiuHemal Shah

Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs. Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Click-Through Rate Prediction Amazon Wide & Deep AUC 0.8637 # 5
Click-Through Rate Prediction Bing News Wide & Deep AUC 0.8377 # 2
Click-Through Rate Prediction Bing News Wide & Deep Log Loss 0.2668 # 2
Click-Through Rate Prediction Company* Wide & Deep (FM & DNN) AUC 0.8661 # 6
Click-Through Rate Prediction Company* Wide & Deep (FM & DNN) Log Loss 0.02640 # 6
Click-Through Rate Prediction Company* Wide & Deep (LR & DNN) AUC 0.8673 # 3
Click-Through Rate Prediction Company* Wide & Deep (LR & DNN) Log Loss 0.02634 # 3
Click-Through Rate Prediction Criteo Wide & Deep (FM & DNN) AUC 0.7850 # 8
Click-Through Rate Prediction Criteo Wide & Deep (FM & DNN) Log Loss 0.45382 # 7
Click-Through Rate Prediction Criteo Wide & Deep (LR & DNN) AUC 0.7981 # 5
Click-Through Rate Prediction Criteo Wide & Deep (LR & DNN) Log Loss 0.46772 # 9
Click-Through Rate Prediction Dianping Wide & Deep AUC 0.8361 # 4
Click-Through Rate Prediction Dianping Wide & Deep Log Loss 0.3364 # 3
Click-Through Rate Prediction MovieLens 20M Wide & Deep AUC 0.7304 # 5