Table 4 Selected hyperparameters of the convolution neural network model with multispectral data.

From: Hyperspectral signature-band extraction and learning: an example of sugar content prediction of Syzygium samarangense

Spectral range

400–1700 nm

400–1000 nm

400–700 nm

900–1700 nm

Conv. kernel size

1 × 5

1 × 2

1 × 2

1 × 5

MaxPooling1 size

1 × 2

1 × 2

2 × 1

2 × 2

MaxPolling2 size

1 × 2

2 × 1

2 × 1

2 × 2

FC1 units

2048

1024

1024

1024

FC2 units

512

256

512

512

FC3 units

1

1

1

1

Batch size

32

64

128

32

Dropout rate

0.3

0.3

0.2

0.1

Learning rate

0.0006

0.0006

0.0006

0.001

Back to article page