CNN-LSTM-Based Late Sensor Fusion for Human Activity Recognition in Big Data Networks

Table 2

Hyperparameter settings for the models.

Hyperparameter	Experimented values	Selected value
CNN layers	1-6	3
Kernel size	3,5,7	3
Feature maps	256, 128, 64, 32	128, 64, 64
Pooling size	2,3,4	2
Dropout	0.2, 0.3, 0.4, 0.5	0.2
Optimizer	RMSProp, Adam	Adam
Learning rate	0.0001 to 0.01	0.001
Batch size	32, 64, 128, 256	64