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

[Baptiste Wicht]

There is something really weird. The GMM version trains almost instantly while the HMM takes days...

Here is the HMM version: 

```cpp
    auto hmm = std::make_unique<HMM<GMM>>(1, GMM(64, 9))
    std::vector<arma::mat> images;
    //Fill images
    hmm->Train(images);
```

Here is the GMM version: 

```cpp
    auto gmm = std::make_unique<GMM>(64, 9);
    std::vector<arma::mat> images;
    //Fill images
    gmm->Train(images[0]);
    for(std::size_t i = 1; i < images.size(); ++i){
        gmm->Train(images[i], 1, true);
    }
```

Either there is something wrong with this code and I don't use GMM/HMM correctly or there is something definitely wrong in the library. 

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/issues/550#issuecomment-193754632
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160308/b64fb45c/attachment.html>