*Client Selection for Federated Learning With Non-IID Data in Mobile Edge Computing*
https://ieeexplore.ieee.org/document/9345723

使用local modelweight與global使用**auxiliary IID dataset**的**auxiliary model weight**計算兩者之間的weight different(其他人是用centralized model來比較)，並以此作為判斷non-iid data程度，接著提高”有較低non-iid”程度local model被選擇參加訓練的程度
(演算法提到**randomly chosen clients**和**the final selected clients**，沒仔細看?)