“Lantian Li 2016-02-22”版本间的差异
来自cslt Wiki
（以“* Weekly Report :* Max-margin metric learning task: : modify the cost function: using sum cost instead of mean cost. : make some experiments based on DNNs tran...”为内容创建页面） |
|||
| 第1行： | 第1行： | ||
* Weekly Report | * Weekly Report | ||
| − | :* Max-margin metric learning task: | + | :* Max-margin metric learning task:[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=lilt&step=view_request&cvssid=485 here] |
: modify the cost function: using sum cost instead of mean cost. | : modify the cost function: using sum cost instead of mean cost. | ||
| 第7行： | 第7行： | ||
: make some experiments based on DNNs transform instead of the simple linear transform. | : make some experiments based on DNNs transform instead of the simple linear transform. | ||
| − | :* Pair-wised speaker vector training. | + | :* Pair-wised speaker vector training.[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=lilt&step=view_request&cvssid=510 here] |
:* Deep speaker embedding. | :* Deep speaker embedding. | ||
2016年2月22日 (一) 12:27的最后版本
- Weekly Report
- Max-margin metric learning task:here
- modify the cost function: using sum cost instead of mean cost.
- make some experiments based on DNNs transform instead of the simple linear transform.
- Pair-wised speaker vector training.here
- Deep speaker embedding.
- code the first version, but it still has lots of bugs. Need Help!
- To Do next week
- max-margin metric learning --> sum cost + DNNs.
- Pair-wised speaker vector training --> Hold.
- Deep speaker embedding --> the CNN structure.
