“Xuewei Zhang 15-07-07”版本间的差异
来自cslt Wiki
| 第7行： | 第7行： | ||
This week: | This week: | ||
1.train RNN+DNN | 1.train RNN+DNN | ||
| − | 2.train clean DNN using knowledge tranfer with 50h and 100h Uythur dataset | + | 2.train clean DNN using knowledge tranfer with 50h and 100h Uythur dataset rspectively |
2015年7月10日 (五) 06:19的最后版本
Last week: 1.train chinglish and chinese language vector 2.train DNN model using knowledge transfer with different hard target rate 3.train clean DNN using knowledge tranfer with 300h Uythur dataset 4.train a model with the concat of 6000h Chinese DNN and one layer RNN using 20h Uythur dataset
This week: 1.train RNN+DNN 2.train clean DNN using knowledge tranfer with 50h and 100h Uythur dataset rspectively
