“Zhiyuan Tang 2016-01-04”版本间的差异
来自cslt Wiki
（以“ Last week: 1. mpe after CTC on WSJ is on the way (a little slow), code rivised ([https://github.com/tzyll/kaldi branch: ctc_customized_20151223]), 2. results of "...”为内容创建页面） |
|||
| 第4行： | 第4行： | ||
1. mpe after CTC on WSJ is on the way (a little slow), code rivised ([https://github.com/tzyll/kaldi branch: ctc_customized_20151223]), | 1. mpe after CTC on WSJ is on the way (a little slow), code rivised ([https://github.com/tzyll/kaldi branch: ctc_customized_20151223]), | ||
| − | 2. results of "chain" on WSJ seems good, mpe has been submitted. | + | 2. results of "chain" on WSJ seems good[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=tangzy&step=view_request&cvssid=492], mpe has been submitted. |
2016年1月4日 (一) 01:00的最后版本
Last week:
1. mpe after CTC on WSJ is on the way (a little slow), code rivised (branch: ctc_customized_20151223),
2. results of "chain" on WSJ seems good[1], mpe has been submitted.
This week:
1. supervise the experiments above;
2. reading.
TASKS
End-to-End
- monophone ASR --Zhiyuan
- MPE
- CTC/nnet3/Kaldi
- chain/nnet3/Kaldi
conditioning learning
- language vector into multiple layers --Zhiyuan
- a Chinese paper
- speech rate into multiple layers --Zhiyuan
- verify the code for extra input(s) into DNN
