Xiangyu Zeng 2015-02-03

来自cslt Wiki
2015年2月3日 (二) 01:30Zengxy讨论 | 贡献的版本

(差异) ←上一版本 | 最后版本 (差异) | 下一版本→ (差异)
跳转至: 导航搜索

last week: 1.add dropout to the dae(noise to clean)and didn't get a good result,it showed that dropout isn't good to dae neetwork 2.train some clean to clean, dae model, and found that we shouldn't add splice as we did in noise to clean dae next week: 1.write a paper about speech rate