[mlpack-git] [mlpack/mlpack] Issues of HMM training (#638)

[Ryan Curtin]

Hi Jinqiang,

I took a look, and the HMM code initializes the transition matrix uniformly.  But I think this is a saddle point for the EM optimization, so the optimization gets stuck.  In 3e4f3ca I added an option to `mlpack_hmm_train` called `--random_initialization`, which will start the optimization with uniformly randomly distributed emission and transition matrices.  I think this will fix your issue.  So if you update to the latest git master branch and rebuild, I think this will solve your problem.  Let me know if it doesn't, and we can reopen the issue and I can try again... :)

Thanks,

Ryan

---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/issues/638#issuecomment-219520481
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160516/52ecd25a/attachment.html>