Table 2 Hyper-parameters of the neural network optimized for predicting each localization-based property. Listed are sets of possible values for each as well as the optimal values determined via Bayesian optimization using SigOpt

Number of layers

[2, 3, 4, 5, 6]

5

2

6

Learning rate

\([1,5,10,50,100]*10^{-4}\)

\(5*10^{-3}\)

\(5*10^{-3}\)

\(10^{-3}\)

Number of epochs

[64, 128, 256, 512, 1024, 2048]

2048

2048

256

Batch size

[8, 16, 32, 64]

64

16

32

Size of layer 1

[8, 16, 32, 64]

8

16

64

Size of layer 2

[16, 32, 64, 128, 256]

128

256

256

Size of layer 3

[16, 32, 64, 128, 256]

128

-

128

Size of layer 4

[16, 32, 64, 128, 256]

128

-

256

Size of layer 5

[16, 32, 64, 128]

16

-

32

Size of layer 6

[8, 16, 32, 64]

-

-

16