Lantian Li 2015-10-12
1. Go on my deep speaker embedding tasks:
1). knowledge transfer for i-vector -- working with Zhiyuan Zhang.
Start to modify nnet-training objective function. Using nnet-train-perutt instead of nnet-train-frame-shuff.
2). Start my new task: metric learning using deep neural networks.
The neural networks are trained using dropout.
The results are still worse than the linear metric learning.
2. Prepare my thesis proposal.
1. Go on the task 1.
Complete modifying the nnet-train-perutt code.
Go on tuning the dropout parameter.
2. Complete my thesis proposal report.