We are happy to share that our paper was honored with the ๐๐๐ฌ๐ญ ๐๐ซ๐๐ฌ๐๐ง๐ญ๐๐ญ๐ข๐จ๐ง ๐๐ฐ๐๐ซ๐ at the ๐๐๐ญ๐ก ๐๐ง๐ญ๐๐ซ๐ง๐๐ญ๐ข๐จ๐ง๐๐ฅ ๐๐จ๐ง๐๐๐ซ๐๐ง๐๐ ๐จ๐ง ๐๐๐๐ก๐ข๐ง๐ ๐๐๐๐ซ๐ง๐ข๐ง๐ ๐๐๐๐ก๐ง๐จ๐ฅ๐จ๐ ๐ข๐๐ฌ (๐๐๐๐๐ ๐๐๐๐) in Helsinki, Finland.
Weโre very grateful for this token of appreciation! At the conference, we had the opportunity to present our recent work, โ๐ ๐๐๐๐ซ๐๐ญ๐๐ ๐๐๐๐ซ๐ง๐ข๐ง๐ ๐๐ข๐ญ๐ก ๐๐ง๐๐ข๐ฏ๐ข๐๐ฎ๐๐ฅ๐ข๐ณ๐๐ ๐๐ซ๐ข๐ฏ๐๐๐ฒ ๐๐ก๐ซ๐จ๐ฎ๐ ๐ก ๐๐ฅ๐ข๐๐ง๐ญ ๐๐๐ฆ๐ฉ๐ฅ๐ข๐ง๐ โ, authored by ๐๐ฎ๐๐๐ฌ ๐๐๐ง๐ ๐, ๐๐ฅ๐ ๐๐จ๐ซ๐๐ก๐๐ซ๐๐ญ, ๐๐ง๐ ๐๐ซ๐ก๐๐ซ๐ ๐๐๐ก๐ฆ. In this paper, we explored how individualized differential privacy can be applied to federated learning clients by adjusting their sampling rates during training. We demonstrated that this approach further improves performance compared to other methods.
