140901 - Mengyuan Zhao

来自cslt Wiki
跳转至: 导航搜索
  • sinovoice AM training:
  1. AM v3.0.S training, 1 iter done.
  2. Test all current AMs in unified configure (beam9, max-actrive5000,acoustic-scale0.10), and collect the results.
  • Reading some papers of convolutional nnet.