Algorithm 1 FederatedAveraging. The K clients are indexed by k, B is the local minibatch size, E is the number of local epochs, and $\eta$ is the learning rate.

Service executes: $w_0\leftarrow$ random initialization for each round t = 1,2,... do     $S_t\leftarrow$ (random subset of $max(C\times K,1)$ clients)     for each client $k\in S_t$ in parallel do         $w_{t+1}^k\leftarrow$ ClientUpdate(k,$w_t$)     end for     $w_{t+1}\leftarrow {\sum }_{k=1}^K\frac{n_k}{n}{w}_{t+1}^k$ end for Client update($k,w$): Split local dataset in B ($\frac{B}{n_k}$ batches of size B) for epoch $e\in \left[1,E\right]$ do     for batch $b\in B$ do         $w\leftarrow w-\eta ▿\ell (w;b)$     end for end for return w to server