Research Article
Early Diagnosis of Retinal Blood Vessel Damage via Deep Learning-Powered Collective Intelligence Models
Table 3
Hyperparameter setting for TDCN-PSO search.
| Category | Hyperparameter | Value |
| Particle swarm optimization | Number of runs () | 5 | Number of iterations () | 12 | Swarm size | 20 | Cg | 0.5 | CNN architecture initialization | Minimum number of outputs from a Conv layer | 3 | Maximum number of outputs from a Conv layer | 256 | Minimum number of neurons in a FC layer | 1 | Maximum number of neurons in a FC layer | 300 | Minimum size of a Conv kernel | | Maximum size of a Conv kernel | | Minimum number of layers | 3 | Maximum number of layers | 20 | Dropout rate | 0.5 | Training | No. of epochs for the global best | 100 | No. of epochs for particle evaluation | 1 | Bath normalize layer outputs | Yes | Probability | Probability of convolutional layer | 0.7 | Probability of pooling layer | 0.15 | Probability of fully connected layer | 0.15 |
|
|