Table 3 The hyper-parameters of the proposed method.
From: Predicting road traffic accident severity from imbalanced data using VAE attention and GCN
Network | Hyper-parameters | Values | Activation function |
|---|---|---|---|
VAE-attention | Optimizer | Adam | – |
Sliding window and steps L, k | 5, 1 | – | |
Input Layer | 9 | – | |
FC layer | 64 | Swish | |
FC layer | 32 | Swish | |
Flatten | – | – | |
Attention layer | 32 | softmax | |
FC layer | \(\mu\) | – | |
FC layer | \(\Sigma\) | – | |
FC layer | 32 | Swish | |
FC layer | 64 | Swish | |
Output Layer | 9 | – | |
GCN | Threshold \(\varepsilon\) | 2.5 | – |
Edge index | (2,num_edges) | – | |
Input Graph | (N, 9) | – | |
GCNConv | (N,16) | Swish | |
Dropout | (N,16) | Swish | |
GCNConv | (N,4) | Log-softmax |
