“140803 - Xuewei Zhang”版本间的差异
来自cslt Wiki
（以内容“Last week: 1.optimize the parameter of Lasso 2.insert one layer or more layers adaptation based on Xent model This week: 1.When the first reverberation is...”创建新页面） |
|||
| 第1行： | 第1行： | ||
| − | Last week: 1.optimize the parameter of Lasso | + | Last week: |
| − | + | 1.optimize the parameter of Lasso | |
| + | 2.insert one layer or more layers adaptation based on Xent model | ||
| − | This week: 1.When the first reverberation is only considered ,that is N=1,the error rate is obtained. | + | This week: |
| − | + | 1.When the first reverberation is only considered ,that is N=1,the error rate is obtained. | |
| − | + | 2.reduce training sets,compute the mean value of the every franme alpha, and the error rate is obtained. | |
| + | 3.apply the Lasso algorithm to Fbank | ||
2014年8月5日 (二) 01:50的最后版本
Last week: 1.optimize the parameter of Lasso 2.insert one layer or more layers adaptation based on Xent model
This week: 1.When the first reverberation is only considered ,that is N=1,the error rate is obtained. 2.reduce training sets,compute the mean value of the every franme alpha, and the error rate is obtained. 3.apply the Lasso algorithm to Fbank
