Table 1 Table of notations.
From: Dynamic weight clustered federated learning for IoT DDoS attack detection
FL components | |
|---|---|
\(N\) | Number of IoT device in the FL system |
\({\text{D}}_{\text{i}},|{\text{D}}_{\text{i}}|\) | The dataset and its size on IoT device i |
\({\upxi }_{\text{k}}\) | Data points consist of pair of features and response |
\(\text{l}({\upphi }_{\text{c}},{\text{D}}_{\text{i}})\) | loss function that measures the performance of the cluster centroid \({\upphi }_{\text{c}}\) on the local data \({\text{D}}_{\text{i}}\) |
\({\upnu }_{\text{i}}(\text{t})\) | The learning rate for Client i in Iteration t |
\(\text{T}\) | Number of local update steps |
Clustering components | |
|---|---|
\(C\) | Number of clusters |
\({\phi }_{\text{i}}\) | Centroid or the average model parameters for cluster \(C\) |
\({s}_{\text{i}}{,}_{\text{k}}\in {\mathbb{R}}^{\text{n}*\text{c}}\) | Binary assignment variable indicating whether client \(i\) is assigned to cluster \(C\), where \({s}_{\text{i}}{,}_{c}=1\) if \(\text{i}\in \text{C}\) else \({s}_{\text{i},\text{c}}=0\) |
\({||\upomega }_{\text{i}}- {\phi }_{\text{i}}|{\left. \right|}_{2}^{2}\) | This represents the squared Euclidean distance between the model parameters of IoT device \(i\) and the centroid of cluster \({\upphi }_{c}\) |
\({\beta }_{\text{i}}\) | The importance weight of IoT device i in Cluster c and \({\sum }_{\text{i}\in \text{k}}{\beta }_{\text{i}}=1\) |
