Table 5 Experimental results with different values of the hyperparameter \(\lambda\)
From: Three-stage training and orthogonality regularization for spoken language recognition
ID | Strategy | \(\lambda\) | channel-test | noisy-test | ||
|---|---|---|---|---|---|---|
| Â | Â | Â | Cavg% | EER% | Cavg% | EER% |
0 | asr-mt-lid-F | 0 | 7.25 | 12.76 | 4.93 | 6.19 |
1 | asr-mt-lid-F | 0.01 | 5.06 | 9.23 | 2.10 | 3.17 |
2 | asr-mt-lid-F | 0.1 | 4.48 | 8.18 | 1.78 | 3.34 |
3 | asr-mt-lid-F | 1 | 5.10 | 9.43 | 2.48 | 5.07 |
4 | asr-mt-lid | 0 | 5.10 | 9.58 | 2.00 | 5.34 |
5 | asr-mt-lid | 0.01 | 5.50 | 10.23 | 1.99 | 5.67 |
6 | asr-mt-lid | 0.1 | 5.30 | 10.66 | 2.02 | 5.74 |
7 | asr-mt-lid | 1 | 4.03 | 8.12 | 2.27 | 5.68 |