Research Article
Pruning Growing Self-Organizing Map Network for Human Physical Activity Identification
Table 4
Training performance of different SOM models
| Model | Hyperparameter | Accuracy (%) | Inference time (ms) | Number of neurons |
| SOM | 20 20 | 79.269 | 0.071 | 400 | 25 25 | 84.405 | 0.110 | 625 | 30 30 | 87.596 | 0.261 | 900 | GSOM | SF = 0.75 | 87.713 | 0.294 | 1044 | SF = 0.80 | 88.027 | 0.315 | 1155 | SF = 0.85 | 88.272 | 0.527 | 1254 | PGSOM (ours) | SF = 0.75, RF = 1.25 | 88.644 | 0.127 | 731 | SF = 0.75, RF = 1.50 | 88.994 | 0.161 | 797.1 | SF = 0.80, RF = 1.25 | 89.401 | 0.265 | 864 | SF = 0.80, RF = 1.50 | 90.120 | 0.262 | 856.1 | SF = 0.85, RF = 1.25 | 89.809 | 0.261 | 864 | SF = 0.85, RF = 1.50 | 89.762 | 0.283 | 929 |
|
|