Using CNN Saliency Maps and EEG Modulation Spectra for Improved and More Interpretable Machine Learning-Based Alzheimer’s Disease Diagnosis

Table 2

CNN hyper-parameter tuning details.

Hyper-parameters	Range explored	Chosen
Kernel_size	(3.3), (5.5), (7.7), (9.9)	(3,3)
Regularizers.l2	(1e − 2), (1e − 4)	(1e − 2)
Dropout	(65%, 75%, 85%, 90%)	85%
Optimizer	Adam, Nadam, Adagrad, Adamax	Nadam
Learning rate	(0.01, 0.001, 0.0001)	0.0001
Batch_size	4, 8, 32, 64	4
Epochs	20, 30, 40, 50	50