1: \begin{abstract}

2: Existing works in federated learning (FL) often assume an ideal system with either full client or uniformly distributed client participation. 

3: However, in practice, it has been observed that some clients may never participate in FL training (aka incomplete client participation) due to a myriad of system heterogeneity factors.

4: A popular approach to mitigate impacts of incomplete client participation is the server-assisted federated learning (SA-FL) framework, where the server is equipped with an auxiliary dataset.

5: However, despite SA-FL has been empirically shown to be effective in addressing the incomplete client participation problem, there remains a lack of theoretical understanding for SA-FL.

6: Meanwhile, the ramifications of incomplete client participation in conventional FL are also poorly understood.

7: These theoretical gaps motivate us to rigorously investigate SA-FL.

8: Toward this end, we first show that conventional FL is {\em not} PAC-learnable under incomplete client participation in the worst case.

9: Then, we show that the PAC-learnability of FL with incomplete client participation can indeed be revived by SA-FL, which theoretically justifies the use of SA-FL for the first time.

10: Lastly, to provide practical guidance for SA-FL training under {\em incomplete client participation}, we propose the \alg (\ul{s}erver-\ul{a}ssisted \ul{f}ederated \ul{a}ve\ul{r}ag\ul{i}ng) algorithm that enjoys the same linear convergence speedup guarantees as classic FL with ideal client participation assumptions, offering the first SA-FL algorithm with convergence guarantee.

11: Extensive experiments on different datasets show \alg significantly improves the performance under incomplete client participation.

12: \end{abstract}

13: