A Fortran-Keras Deep Learning Bridge for Scientific Computing

Table 1

Hyperparameter space.

Name	Options	Parameter type
Batch normalization	[yes, no]	Choice
Dropout	[0, 0.25]	Continuous
Leaky ReLU coefficient	[0–0.4]	Continuous
Learning rate	[0.00001–0.01]	Continuous (log)
Nodes per layer	[125, 256, 512]	Discrete
Number of layers	[4–11]	Discrete
Optimizer	[Adam, RMSProp, SGD]	Choice