Jiangchao Yao, Feng Wang, Xichen Ding, Shaohu Chen, Bo Han...
TL;DR本研究提出一种基于 meta 控制器和因果透视的样本构建方法以解决设备层和云端层之间协同推荐的问题，并在工业推荐场景中展示了 meta 控制器的潜力。
Abstract
on-device machine learning enables the lightweight deployment of
recommendation models in local clients, which reduces the burden of the
cloud-based recommenders and simultaneously incorporates more real-time use