Table 2.

Hyperparameters used in feature extraction.

Momentum	Initial learning rate	Decay	Number of epochs per decay	Weight decay rate	End learning rate	Batch size
0.9	0.01	0.95	2	0.00004	0.00001	32