version

来自「Matlab 马尔科夫计算工具包」· 代码 · 共 25 行

What's new in version 3.x:

1. You can initialise the state transition matrices (hmm.P) before
you do hmmtrain.

2. A demonstration, demar, showing an HMM with an AR
observation model applied to a section of sleep spindle data.

3. More help text.

4. Includes a weighted EM algorithm for training Gaussian Mixture models where each data point is weighted by an amount `gamma'.

5. A feature allowing you to update only selected observation models during training (ie observation models for some states stay the same).

6. Version 3.1 contains the files: arembed.m, embed.m and arwls.m that should've been distributed with 3.0.

7. Version 3.2 also contains: init_ar.m (and dar.m) necessary for initialising 
AR vectors in HMM-AR model

8. Version 3.2 also contains another demo: demar2.m which uses init_ar.m

9. Version 3.2 also contains a file which allows for diagonal/full
covariance matrices