Learning Parameters

Learning parameters are stored and handled in the base_optimiser_type derived type. For a brief overview of what is available in that derived type, please refer to that page.

For optimisation methods, there exist the following methods:

• None
• SGD (momentum-based)
• RMSprop
• Adam
• AdaGrad

For regularisation methods, there exist the following methods in the library:

• None
• L1
• L2
• L1L2

For learning rate decay methods, there exist the following methods in the library:

• None
• Exponential
• Step
• Inverse