Skip to main content

Table 1 Parameter settings for neural-network-based methods

From: Data assessment and prioritization in mobile networks for real-time prediction of spatial information using machine learning

Batch size

50

Epochs

30

Optimizer

Adam [24]

Learning rate

0.001 [24]

Loss function

Mean absolute error

MLP

Input (no. of units=1440)

 

Batch normalization

 

Dense (no. of units=512)

 

Batch normalization

 

Activation (ReLU)

 

Dense (no. of units=256)

 

Batch normalization

 

Activation (ReLU)

 

Dense (no. of units=144)

3D-CNN

Input (shape=(10,12,12))

 

Conv3D (filters=32, kernel_size=(3,3,3),

 

padding=‘same’)

 

Batch normalization

 

Activation (ReLU)

 

AveragePooling3D (pool_size=(2,2,2),

 

padding=‘same’)

 

Conv3D (filters=32, kernel_size=(3,3,3),

 

padding=‘same’)

 

Batch normalization

 

Activation (ReLU)

 

Dense (144)

LSTM

Input (shape=(10,144))

 

LSTM (no. of units=128)

 

LSTM (no. of units=64)

 

Dense (no. of units=144)

Other parameters

Keras default settings