CTR Prediction on Alibaba Dataset (test)
0.6541AUCDIEN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DIENcomponents=GRU + AUGRU + auxiliary loss2018.09 | 0.6541 | — | — | |
| BaseModel + GRU + AUGRUcomponents=BaseModel + GRU + Attentional Update Gate GRU2018.09 | 0.6493 | — | — | |
| Two layer GRU Attentionlayers=two-layer GRU, mechanism=attention layer2018.09 | 0.6457 | — | — | |
| DINmechanism=attention2018.09 | 0.6428 | — | — | |
| Wide & Deeparchitecture=deep model + linear wide model2018.09 | 0.6362 | — | — | |
| PNNinteraction=product layer2018.09 | 0.6353 | — | — | |
| BaseModelpooling=sum pooling2018.09 | 0.635 | — | — | |
| DIN with MBA Reg. and DiceActivation function=Dice, Regularization=Mini-batch Aware (MBA) Regularization2017.06 | 0.6083 | 11.65 | — | |
| DIN with MBA Reg.Activation function=PReLU, Regularization=Mini-batch Aware (MBA) Regularization2017.06 | 0.606 | 9.28 | — | |
| DIN with DiceActivation function=Dice, Regularization=Dropout2017.06 | 0.6044 | 7.63 | — | |
| DIN ModelActivation function=PReLU, Regularization=Dropout2017.06 | 0.6029 | 6.08 | — | |
| DeepFMActivation function=PReLU, Regularization=Dropout2017.06 | 0.5993 | 2.37 | — | |
| PNNActivation function=PReLU, Regularization=Dropout2017.06 | 0.5983 | 1.34 | — | |
| Wide&DeepActivation function=PReLU, Regularization=Dropout2017.06 | 0.5977 | 0.72 | — | |
| BaseModelActivation function=PReLU, Regularization=Dropout2017.06 | 0.597 | 0 | — | |
| LR2017.06 | 0.5738 | -23.92 | — | |
| CentralSGDclients per round (K)=N/A (Centralized)2021.09 | — | — | 265 | |
| FedSubAvgclients per round (K)=1002021.09 | — | — | 610 |