Why are there Topic Concentration Iterations when you disable fitting topic concentration in fitlda?

1 次查看（过去 30 天）
Stephen Bruestle
Stephen Bruestle 2019-1-17
评论： Stephen Bruestle 2019-1-17
When I run fitlda with 'FitTopicConcentration' set to false, I still get `Topic Concentration Interatons'.
For example:
mdl = fitlda(bag,numTopics,'Verbose',1,'InitialTopicConcentration',50,'FitTopicConcentration',false,'LogLikelihoodTolerance',0,'IterationLimit',20);
gets:
Initial topic assignments sampled in 1.44622 seconds.
=====================================================================
| Iteration | Time per | Relative | Training | Topic |
| | iteration | change in | perplexity | concentration |
| | (seconds) | log(L) | | iterations |
=====================================================================
| 0 | 0.90 | | 6.715e+02 | 0 |
| 1 | 1.79 | 3.3759e-02 | 5.429e+02 | 0 |
| 2 | 1.70 | 1.2228e-02 | 5.031e+02 | 0 |
| 3 | 1.58 | 2.1513e-03 | 4.965e+02 | 0 |
| 4 | 1.57 | 6.2666e-04 | 4.945e+02 | 0 |
| 5 | 1.55 | 3.3881e-04 | 4.935e+02 | 0 |
| 6 | 1.60 | 2.5182e-04 | 4.927e+02 | 0 |
| 7 | 1.55 | 1.5109e-04 | 4.923e+02 | 0 |
| 8 | 1.62 | 2.2085e-04 | 4.916e+02 | 0 |
| 9 | 1.64 | 2.1745e-04 | 4.909e+02 | 0 |
| 10 | 1.66 | 1.0178e-04 | 4.906e+02 | 0 |
| 11 | 1.86 | 1.1432e-04 | 4.903e+02 | 4 |
| 12 | 1.79 | 6.5833e-04 | 4.883e+02 | 3 |
| 13 | 1.77 | 4.3774e-04 | 4.870e+02 | 3 |
| 14 | 1.70 | 4.8030e-04 | 4.855e+02 | 3 |
| 15 | 1.73 | 4.1302e-04 | 4.843e+02 | 3 |
| 16 | 1.69 | 1.9747e-04 | 4.837e+02 | 3 |
| 17 | 1.73 | 2.5513e-04 | 4.829e+02 | 3 |
| 18 | 1.83 | 2.4952e-04 | 4.822e+02 | 3 |
| 19 | 1.63 | 9.3627e-05 | 4.819e+02 | 2 |
| 20 | 1.66 | 1.0886e-04 | 4.816e+02 | 2 |
=====================================================================
Why is the rightmost column not all zeros? What does "topic concentration iterations" mean?
  1 个评论
Stephen Bruestle
Stephen Bruestle 2019-1-17
Note, when I run:
mdl = fitlda(bag,numTopics,'Verbose',1,'InitialTopicConcentration',50,'FitTopicConcentration',true,'LogLikelihoodTolerance',0,'IterationLimit',20);
I get:
Initial topic assignments sampled in 1.31452 seconds.
=====================================================================================
| Iteration | Time per | Relative | Training | Topic | Topic |
| | iteration | change in | perplexity | concentration | concentration |
| | (seconds) | log(L) | | | iterations |
=====================================================================================
| 0 | 0.59 | | 6.715e+02 | 50.000 | 0 |
| 1 | 1.59 | 3.2014e-02 | 5.487e+02 | 50.000 | 0 |
| 2 | 1.57 | 1.1809e-02 | 5.098e+02 | 50.000 | 0 |
| 3 | 1.57 | 2.9063e-03 | 5.006e+02 | 50.000 | 0 |
| 4 | 1.68 | 7.4965e-04 | 4.983e+02 | 50.000 | 0 |
| 5 | 1.55 | 4.2984e-04 | 4.970e+02 | 50.000 | 0 |
| 6 | 1.61 | 2.2000e-04 | 4.963e+02 | 50.000 | 0 |
| 7 | 1.57 | 1.3672e-04 | 4.959e+02 | 50.000 | 0 |
| 8 | 1.55 | 3.0094e-04 | 4.950e+02 | 50.000 | 0 |
| 9 | 1.47 | 1.7135e-04 | 4.944e+02 | 50.000 | 0 |
| 10 | 1.53 | 1.9109e-04 | 4.939e+02 | 50.000 | 0 |
| 11 | 3.51 | 5.6307e-05 | 4.937e+02 | 8.503 | 47 |
| 12 | 2.40 | 4.8823e-02 | 3.699e+02 | 6.363 | 22 |
| 13 | 2.05 | 7.0746e-03 | 3.548e+02 | 6.023 | 13 |
| 14 | 2.01 | 1.7874e-03 | 3.511e+02 | 5.897 | 9 |
| 15 | 2.43 | 1.2725e-03 | 3.485e+02 | 5.775 | 9 |
| 16 | 1.89 | 1.2613e-03 | 3.460e+02 | 5.625 | 10 |
| 17 | 1.88 | 1.1882e-03 | 3.436e+02 | 5.512 | 9 |
| 18 | 1.90 | 1.2557e-03 | 3.411e+02 | 5.387 | 9 |
| 19 | 1.89 | 9.4163e-04 | 3.392e+02 | 5.307 | 7 |
| 20 | 1.94 | 8.2492e-04 | 3.376e+02 | 5.217 | 8 |
=====================================================================================

请先登录，再进行评论。

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Modeling and Prediction 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by