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

Mon May 16 00:59:23 EDT 2016

Hi rcurtln,

Sorry for the late reply.
Here's the training set I used in the beginning.
data.txt is the file that contains all other observation sequences. And other files contain three or four normalized observations with 6 DOF.

"data.txt"
data11.txt
data21.txt
data31.txt
data41.txt
data51.txt

"data11.txt"
1.407188 -0.825524 -0.581664 1.330260 -0.249431 -1.080829
1.064666 0.273809 -1.338475 -1.414204 0.702534 0.711669
1.118510 0.190213 -1.308723 -0.978695 1.373439 -0.394744

"data21.txt"
1.413624 -0.742167 -0.671457 1.393726 -0.489149 -0.904577
1.414185 -0.699330 -0.714855 1.414114 -0.721602 -0.692512
1.413064 -0.755915 -0.657149 -1.414117 0.721405 0.692712

"data31.txt"
1.731182 -0.618199 -0.529350 -0.583632 1.604384 0.056298
-0.682269 -0.978413 1.210370 0.761583 -0.912841 -1.059112
-0.443458 1.726382 -0.642271 -0.640653 1.157285 0.828081
-0.947201 -1.038165 -0.964568 -1.034804 1.004618 0.994753

"data41.txt"
1.731235 -0.530060 -0.585568 -0.615607 1.636155 -0.075307
-0.557711 -1.003137 1.190107 0.789548 -0.970546 -1.009108
-0.423812 1.724491 -0.632598 -0.668081 1.092094 0.901500
-0.935888 -1.057706 -0.977638 -1.011645 0.849429 1.139854

"data51.txt"
1.715604 -0.348052 -0.699034 -0.668518 1.633409 -0.045010
-0.609211 -0.979189 1.245627 0.717635 -0.925878 -1.037384
1.496415 0.177980 -0.432682 -1.241713 1.319591 0.616885
-1.007997 -0.928479 -0.970147 -1.026679 0.923957 1.072869

The output is as follows:
[INFO ] Reading list of training sequences from 'data.txt'.
[INFO ] Adding training sequence from 'data11.txt'.
[INFO ] Loading 'data11.txt' as raw ASCII formatted data.  Size is 6 x 3.
[INFO ] Adding training sequence from 'data21.txt'.
[INFO ] Loading 'data21.txt' as raw ASCII formatted data.  Size is 6 x 3.
[INFO ] Adding training sequence from 'data31.txt'.
[INFO ] Loading 'data31.txt' as raw ASCII formatted data.  Size is 6 x 4.
[INFO ] Adding training sequence from 'data41.txt'.
[INFO ] Loading 'data41.txt' as raw ASCII formatted data.  Size is 6 x 4.
[INFO ] Adding training sequence from 'data51.txt'.
[INFO ] Loading 'data51.txt' as raw ASCII formatted data.  Size is 6 x 4.
[INFO ] 
[INFO ] Execution parameters:
[INFO ]   batch: true
[INFO ]   gaussians: 0
[INFO ]   help: false
[INFO ]   info: ""
[INFO ]   input_file: data.txt
[INFO ]   labels_file: ""
[INFO ]   model_file: ""
[INFO ]   output_model_file: dataout.xml
[INFO ]   seed: 0
[INFO ]   states: 6
[INFO ]   tolerance: 1e-05
[INFO ]   type: gaussian
[INFO ]   verbose: true
[INFO ]   version: false
[INFO ] 
[INFO ] Program timers:
[INFO ]   loading_data: 0.000880s
[INFO ]   total_time: 0.062300s

Then  I used one file contains a list of datas:
-0.049332,0.632684,1.930238
-0.045342,0.653824,1.949537
0.223055,0.607684,1.885569
0.456205,0.482629,1.495054
0.62325,0.454105,1.333082
0.740229,0.468617,1.399701
0.758242,0.476387,1.570806
0.772725,0.504061,1.601133
0.799275,0.537727,1.666062
0.872294,0.580186,1.646654
0.889642,0.612556,1.607951
0.835107,0.669644,1.528674
0.704884,0.683599,1.427598
0.736671,0.727499,1.386765
0.875919,0.762281,1.269978
0.974008,0.842077,0.490348
1.041494,0.898995,0.123952
1.061031,0.943553,0.208335
0.960244,0.917036,1.008331
0.54015,0.698827,1.148918
0.350406,0.631777,1.10327
0.242643,0.615209,1.044297
0.238878,0.624707,1.015939
0.336262,0.713242,1.063636
0.743459,0.792901,1.232002
1.005647,0.869002,1.491689
1.205512,0.929804,1.472736
1.149689,0.895766,1.373127
1.059189,0.759632,1.278505
0.978293,0.743314,1.199621
0.917716,0.755696,1.180213
0.686448,0.741106,1.228388
0.576865,0.679422,1.248577
0.574938,0.675374,1.475239
0.62893,0.690035,1.629235
0.699362,0.786216,1.768085
0.695823,0.876179,1.723242
0.509356,0.969935,1.704582

or discrete datas:
1
7
8
9
4
3
1
7
8
9
4
3
1
7
8
9
4
3

The output of the program is the same and the result models turn out to be similar: uniform matrix and weird emissions.
Thank you.

Jinqiang

---
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-219349141
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160515/0a5c4eb3/attachment.html>