“Zhiyuan Tang 2015-03-23”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
 
第7行: 第7行:
  
 
the CONTEXT into DNN may be not enough, as the central frame is not highlighted and its neighbors should not share the same importance. Analysis of the first-layer weights may give some insight to improve the CONTEXT.  
 
the CONTEXT into DNN may be not enough, as the central frame is not highlighted and its neighbors should not share the same importance. Analysis of the first-layer weights may give some insight to improve the CONTEXT.  
 +
  
 
This week:
 
This week:

2015年3月23日 (一) 13:36的最后版本

Last week:

got a better understanding of ASR with some papers;

did some easy analysis about the weights of DNN already learned from Kaldi to see if some intersting things could be found and more regularization could be added to train a better DNN;

the CONTEXT into DNN may be not enough, as the central frame is not highlighted and its neighbors should not share the same importance. Analysis of the first-layer weights may give some insight to improve the CONTEXT.


This week:

do more analysis about the learned weights;

improve the CONTEXT if possible, at least give it some small changes and contrast its performance with previous results.