1 paper · 2 filters
Yucheng Wu, Yuekui Yang, Hongzheng Li +6
Deploying new architectures in large-scale user response prediction systems incurs high model switching costs due to expensive retraining on massive historical data and performance…