Table 2.
Optimization algorithm and hyperparameter settings for training the CNN.
Sound Event | Speech Command | |
---|---|---|
Optimization algorithm | Adam | Adam |
Initial learn rate | 0.001 | 0.0003 |
Mini batch size | 50 | 128 |
Max epochs | 30 | 25 |
Learn rate drop factor | 0.5 | 0.1 |
Learn rate drop period | 6 | 20 |
L2 regularization | 0.05 | 0.05 |