“2015-05-04”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
第22行: 第22行:
  
 
3. Translate the patent.
 
3. Translate the patent.
 +
 +
[[Lantian Li 15-05-05]]

2015年5月5日 (二) 01:18的版本

Weekly Summary

1. Build the cohort-based DNN classifier with discriminative features.

Exp. results show that the performance of the system with DNN classifier is more unstable compared with SVM classifier .

For DNN classifier, the best performance achieves the EER of 0.90%, while the worst is up to 2.54%.

For SVM classifier, the best performance achieves the EER of 1.48%, while the worst is up to 1.61%.

2. The SUSR paper has been modified.

3. Help the D-ear company: write a patent (completed).

4. Submit my ChinaSIP final paper.

Next Week

1. Go on the task 1.

2. Submit the SUSR paper to IEEE Trans.

3. Translate the patent.

Lantian Li 15-05-05