Table 4 Simulation hyperparameters of traditional models used for comparative analysis.
From: Hybrid quantum enhanced federated learning for cyber attack detection
S.No | Model | Parameter | Range/Type |
|---|---|---|---|
1 | Random forest | Number of trees | 200 |
2 | Max depth | 30 | |
3 | Min samples split | 2 | |
4 | Min samples leaf | 1 | |
5 | CNN | Convolutional layers | 4 |
6 | Filter size | (3, 3) | |
7 | Pooling size | (2, 2) | |
8 | Activation | ReLU | |
9 | Optimizer | Adam | |
10 | Learning rate | 0.001 | |
11 | LSTM | Number of layers | 3 |
12 | Units per layer | 128 | |
13 | Dropout | 0.2 | |
14 | Activation | Tanh | |
15 | Optimizer | RMSprop | |
16 | Learning rate | 0.001 | |
17 | RNN | Number of layers | 3 |
18 | Units per layer | 64 | |
19 | Dropout | 0.3 | |
20 | Activation | Sigmoid | |
21 | Optimizer | Adam | |
22 | Learning rate | 0.0005 | |
23 | FL | Nodes | 10 |
24 | Communication rounds | 200 | |
25 | Optimizer | SGD | |
26 | Learning rate | 0.01 | |
27 | SSTDL | Convolutional layers | 3 |
28 | Activation | ReLU | |
29 | Optimizer | Adam | |
30 | Learning rate | 0.001 | |
31 | Dropout | 0.3 | |
32 | STGNN | Graph layers | 3 |
33 | Attention heads | 8 | |
34 | Node embedding size | 128 | |
35 | Aggregation method | Mean | |
36 | Optimizer | RMSprop | |
37 | Learning rate | 0.0005 | |
38 | Dropout rate | 0.2 |
