Combining Feature Importance and Bilinear feature Interaction for CTR Prediction (FiBiNET)
时间:2019.05作者:TongwenHuang,ZhiqiZhang,JunlinZhangAbstract文中使用Squeeze-ExcitationNetwork(SENET)动态学习特征的重要性,并且使用bilinearfunction学习特征组合ModelStructureSENETLayerSqueeze:用max/averagepooling学习每一维特征的全局信息Excitati