1mlpack_hmm_train(1) User Commands mlpack_hmm_train(1)
2
3
4
6 mlpack_hmm_train - hidden markov model (hmm) training
7
9 mlpack_hmm_train -i string [-b bool] [-g int] [-m unknown] [-l string] [-s int] [-n int] [-T double] [-t string] [-V bool] [-M unknown] [-h -v]
10
12 This program allows a Hidden Markov Model to be trained on labeled or
13 unlabeled data. It support three types of HMMs: discrete HMMs, Gaussian
14 HMMs, or GMM HMMs.
15
16 Either one input sequence can be specified (with --input_file), or, a
17 file containing files in which input sequences can be found (when
18 --input_file and --batch are used together). In addition, labels can be
19 provided in the file specified by --labels_file, and if --batch is
20 used, the file given to --labels_file should contain a list of files of
21 labels corresponding to the sequences in the file given to
22 --input_file.
23
24 The HMM is trained with the Baum-Welch algorithm if no labels are pro‐
25 vided. The tolerance of the Baum-Welch algorithm can be set with the
26 --tolerance option. By default, the transition matrix is randomly ini‐
27 tialized and the emission distributions are initialized to fit the
28 extent of the data.
29
30 Optionally, a pre-created HMM model can be used as a guess for the
31 transition matrix and emission probabilities; this is specifiable with
32 --model_file.
33
35 --input_file (-i) [string]
36 File containing input observations.
37
39 --batch (-b) [bool]
40 If true, input_file (and if passed, labels_file) are expected to
41 contain a list of files to use as input observation sequences
42 (and label sequences).
43
44 --gaussians (-g) [int]
45 Number of gaussians in each GMM (necessary when type is 'gmm').
46 Default value 0.
47
48 --help (-h) [bool]
49 Default help info.
50
51 --info [string]
52 Get help on a specific module or option. Default value ''.
53
54 --input_model_file (-m) [unknown]
55 Pre-existing HMM model to initialize training with. Default
56 value ''.
57
58 --labels_file (-l) [string]
59 Optional file of hidden states, used for labeled training.
60 Default value ''.
61
62 --seed (-s) [int]
63 Random seed. If 0, 'std::time(NULL)' is used. Default value 0.
64
65 --states (-n) [int]
66 Number of hidden states in HMM (necessary, unless model_file is
67 specified). Default value 0.
68
69 --tolerance (-T) [double]
70 Tolerance of the Baum-Welch algorithm. Default value 1e-05.
71
72 --type (-t) [string]
73 Type of HMM: discrete | gaussian | gmm. Default value 'gauss‐
74 ian'.
75
76 --verbose (-v) [bool]
77 Display informational messages and the full list of parameters
78 and timers at the end of execution.
79
80 --version (-V) [bool]
81 Display the version of mlpack.
82
84 --output_model_file (-M) [unknown]
85 Output for trained HMM. Default value ''.
86
88 For further information, including relevant papers, citations, and the‐
89 ory, consult the documentation found at http://www.mlpack.org or
90 included with your distribution of mlpack.
91
92
93
94mlpack-3.0.4 21 February 2019 mlpack_hmm_train(1)