Lantian Li 2016-02-29

From cslt Wiki
Revision as of 00:52, 29 February 2016 by Lilt (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
  • Weekly Report
  • Max-margin metric learning task:here
verify the difference between Fisher / SRE05 / SWC.
deep structure is better than the simple linear transform, and it seems that the mean cost is better than the sum cost.
  • Pair-wised speaker vector
  • Deep speaker embedding.
code the first version, but it still has lots of bugs. Need Help!
  • d-vector nnet3 reproduce.
Data preparation and code modification.