|
|
| 第46行： |
第46行： |
| | |Lantian Li | | |Lantian Li |
| | || | | || |
| − | * | + | * Failed the VAE-based Bayesian scoring. |
| | + | * Review Bayesian hypothesis test. |
| | || | | || |
| − | * | + | * Study on NDA scoring. |
| | || | | || |
| | * | | * |
| People |
This Week |
Next Week |
Task Tracking (DeadLine)
|
| Dong Wang
|
- Establish neural Bayes theory
- Unify the subspace DNF and NDA
- More investigation on neural PCA
|
- More work on neural PCA
- Investigate incremental training algorithm
|
|
| Yunqi Cai
|
- Did some experiments on VAE Bayes score
- Apply NDA to xvector's deep normlization
- Investigate the literature on Deep generative model
|
- keep on literature Investigation
- Give some consistent reults on NDA-xvector deep normlization
|
|
| Zhiyuan Tang
|
- Training NDA for TTS (also used for speaker conversion/speaker recognition).
- A patent tech doc about flow encryptor.
|
|
|
| Lantian Li
|
- Failed the VAE-based Bayesian scoring.
- Review Bayesian hypothesis test.
|
|
|
| Ying Shi
|
- Investigate how to apply mse loss on parallel flow
- experiment about unsupervised sequential flow
|
- train unsupervised sequential flow
- experiment about supervised sequential flow
|
|
| Wenqiang Du
|
|
|
|
| Haoran Sun
|
|
|
|
| Yue Fan
|
- Data collection
- Program processing
|
- Train dnf model with gun&nongun and test
|
|
| Jiawen Kang
|
- Adapting meta learning code to speaker recognition task.
|
|
|
| Ruiqi Liu
|
- Study meta-learning experiments and modify the code.
|
- Continue to study meta-learning experiments.
- Provide ASR training tools for Shengyuan company.
|
|
| Sitong Cheng
|
|
|
|
| Zhixin Liu
|
|
|
|
| Haolin Chen
|
- Cooperate with Shi Ying on experiments
|
|
|
