Skip to main content
. 2020 May 30;10(6):358. doi: 10.3390/diagnostics10060358

Table 4.

Optimal values for the hyperparameters for the custom wide residual network (WRN) and pretrained convolutional neural networks (CNNs) obtained through the randomized grid search (M: momentum, ILR: initial learning rate, and L2: L2-weight decay).

Models Optimal Values
M ILR L2
Custom 0.90 1 × 10−3 1 × 10−5
Pretrained 0.95 1 × 10−3 1 × 10−6