Skip to main content

Table 5 Architectures of our highest-performing MLP and RBF networks

From: Analysis of risk factor domains in psychosis patient health records

Network MLP RBF
Input Layer   
Nodes 512 512
Dropout 0.2 0.2
Activation ReLU ReLU
Hidden Layer   
Nodes 250 700
Dropout 0.5 0.0
Activation ReLU RBF
Output Layer   
Nodes 7 7
Activation Sigmoid Linear
Optimizer Adam Adam
Loss Function Categorical Cross Entropy Mean Squared Error
Training Epochs 60 50
Batch Size 128 128