Table D1. Open in new tab... | Oxford Academic

Table D1.

Open in new tab

Hyperparameters used in a CNN architecture trained using spectra with a higher S/N. These values are selected using a Bayesian optimization algorithm.

	Hyperparameters	Optimized value
Data Input	window size (ws)	285
	cnpix	1
CNN	L2	0.0
Architecture	dropout	0.6
	conv_filter_1	256
	conv_filter_2	256
	conv_filter_3	512
	conv_kernel_1	10
	conv_kernel_2	7
	conv_kernel_3	6
	dense_1	64
	dense_2_ID	512
	dense_2_N	512
	dense_2_z	256
	dense_2_b	128

	Hyperparameters	Optimized value
Data Input	window size (ws)	285
	cnpix	1
CNN	L2	0.0
Architecture	dropout	0.6
	conv_filter_1	256
	conv_filter_2	256
	conv_filter_3	512
	conv_kernel_1	10
	conv_kernel_2	7
	conv_kernel_3	6
	dense_1	64
	dense_2_ID	512
	dense_2_N	512
	dense_2_z	256
	dense_2_b	128

Table D1.

Open in new tab

Hyperparameters used in a CNN architecture trained using spectra with a higher S/N. These values are selected using a Bayesian optimization algorithm.

	Hyperparameters	Optimized value
Data Input	window size (ws)	285
	cnpix	1
CNN	L2	0.0
Architecture	dropout	0.6
	conv_filter_1	256
	conv_filter_2	256
	conv_filter_3	512
	conv_kernel_1	10
	conv_kernel_2	7
	conv_kernel_3	6
	dense_1	64
	dense_2_ID	512
	dense_2_N	512
	dense_2_z	256
	dense_2_b	128

	Hyperparameters	Optimized value
Data Input	window size (ws)	285
	cnpix	1
CNN	L2	0.0
Architecture	dropout	0.6
	conv_filter_1	256
	conv_filter_2	256
	conv_filter_3	512
	conv_kernel_1	10
	conv_kernel_2	7
	conv_kernel_3	6
	dense_1	64
	dense_2_ID	512
	dense_2_N	512
	dense_2_z	256
	dense_2_b	128