“Zhiyuan Tang 2015-03-23”版本间的差异
来自cslt Wiki
（以“ 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...”为内容创建页面） |
|||
| 第1行： | 第1行： | ||
Last week: | Last week: | ||
| + | |||
got a better understanding of ASR with some papers; | 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; | 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. | 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: | ||
| + | |||
do more analysis about the learned weights; | 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. | improve the CONTEXT if possible, at least give it some small changes and contrast its performance with previous results. | ||
2015年3月23日 (一) 13:35的版本
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.
