[mlpack-git] [mlpack] How to speed up HMM training ? (#550)

Baptiste Wicht notifications at github.com
Tue Mar 8 02:01:01 EST 2016


I just got the results this morning. It took almost 53 hours to train the GMM :(

> Have you compiled mlpack without debugging symbols? 

I'm using the release 2.0.1 built from sources with the default options, so I don't think there are debugging in it. 

> And I assume you're compiling your program with -O3 and without any debugging symbols or anything.

The program is compiled with -O2 -g -march=native with g++-4.9.3

> I think maybe to figure out what is going on here, you should try removing the HMM layer and just train a GMM on its own on the data

I haven't found out how to train a GMM with multiple sequences. Do I have to loop and call Train several times ? 

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/issues/550#issuecomment-193635112
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160307/7e2f85a7/attachment-0001.html>


More information about the mlpack-git mailing list