Table 1 Proposed pseudocode (server end).
From: A fused weighted federated learning-based adaptive approach for early-stage drug prediction
Proposed Algo steps |
|---|
Begin |
Set ₀(,) & ₀(,), where ₀(,) is the weight of nodes between the input layer and the hidden layers at the server end, ₀(, fml) |
For each cycle k from 1 to K, do: |
a. Randomly, from the general population of clients described by η forState Sk of clients |
b. For each client l in Sk, in parallel, do the following: |
i. Perform Client Training on the new weights wn(k + 1) and vn(k + 1) |
c. Aggregate the updated weights from all clients: |
i. _(G, fml)^k = (1 / (Σ_n ε η Σ_(n = 1)^N S_n/S * wn(k + 1))) |
ii. _(G, fml)^k = (1 / (Σ_n ε η Σ_(n = 1)^N (S_n/S * vn(k + 1)))) |
Stop |
End For |
End For |
