“Zhiyong Zhang 2015-12-21”版本间的差异
来自cslt Wiki
| 第1行： | 第1行： | ||
* Weekly Report | * Weekly Report | ||
# Analysed the MPE divergence problem when the context is 10. The reason may be ascribed to the Over-fitting. | # Analysed the MPE divergence problem when the context is 10. The reason may be ascribed to the Over-fitting. | ||
| − | Large data-set may weaken the over-fitting phenomenon. | + | # Large data-set may weaken the over-fitting phenomenon. |
* Target Next Week | * Target Next Week | ||
# Read some paper on CTC and Neural Turing Machines; | # Read some paper on CTC and Neural Turing Machines; | ||
# Read kaldi CTC related scripts to familiar with the framework. | # Read kaldi CTC related scripts to familiar with the framework. | ||
2015年12月21日 (一) 14:03的最后版本
- Weekly Report
- Analysed the MPE divergence problem when the context is 10. The reason may be ascribed to the Over-fitting.
- Large data-set may weaken the over-fitting phenomenon.
- Target Next Week
- Read some paper on CTC and Neural Turing Machines;
- Read kaldi CTC related scripts to familiar with the framework.
