Table 1.

Pytorch CNN settings used for all subsequent tests.

Training/Test Data/Number of Spectra	80%/20%/1603
Number of convolution layers	3
Kernel sizes	3, 3, 3
Number of descriptors	8, 16, 32
Maximum pooling sizes	3, 3, 3
Number of MLP hidden layers	3
Sizes of hidden layers	256, 128, 64
Optimizer	AdamW
Activation function	Leaky ReLU
Learning rate	9 × 10−7
Loss function	Cross entropy