Difference between revisions of "Lantian Li 2016-08-22"

From cslt Wiki
Jump to: navigation, search
Line 5: Line 5:
 
:* Deep speaker embedding (Random sample training).
 
:* Deep speaker embedding (Random sample training).
  
Make the d-vector baseline on Theano.
+
  Make the d-vector baseline on Theano.
  
 
:* Local training.  
 
:* Local training.  
  
Data preparation -- Done.
+
  Data preparation -- Done.
  
T-matrix training.
+
  T-matrix training.
  
 
:* Deep speaker embedding ( LSTM + LR/Margin ) -- hold.
 
:* Deep speaker embedding ( LSTM + LR/Margin ) -- hold.

Revision as of 03:38, 22 August 2016

  • Weekly Report
  • OLE evaluation task. -- Done
  • Deep speaker embedding (Random sample training).
  Make the d-vector baseline on Theano.
  • Local training.
  Data preparation -- Done.
  T-matrix training.
  • Deep speaker embedding ( LSTM + LR/Margin ) -- hold.
  • This Week
  • Go on three tasks.