HMMUnsupervised Arguments - Teradata Vantage

Teradata® Vantage Machine Learning Engine Analytic Function Reference

Teradata Vantage
Release Number
May 2019
English (United States)
Last Update
Product Category
Teradata Vantage
[Optional] Specify the name of the input column that contains the model attribute. If you specify this argument, model_column must match a model_key in the PARTITION BY clause. The values in the column can be either integers or strings.
Specify the name of the input column that contains the sequence attribute. The sequence_column must be a sequence attribute in the PARTITION BY clause and contain more than two observation symbols.
Specify the name of the input column that contains the observed symbols. The function scans the input table to find all possible observed symbols.
Observed symbols are case-sensitive.
Specify the number of hidden states.
The number of hidden states can influence model quality and performance, so choose the number appropriately.
[Optional] Specify the number of iterations that the training process runs before the function completes.
Default: 10

[Optional] Specify the threshold value that determines the convergence of HMM training. If the parameter value difference is less than the threshold, the training process converges.

Default behavior: Only MaxIterNum determines when the training process converges.

[Optional] Specify the name of the input column whose value determines whether the function skips the row.
[Optional] Specify the method for creating the initial parameters for the initial state probabilities, state transition probabilities, and observation emission probabilities:
Option Description
'random' (Default) Initial parameters are based on uniform distribution.
The seed is meaningful only with 'random'. Specify the random seed the algorithm uses for repeatable results (for more information, see Nondeterministic Results).
'flat' Probabilities are equal. Each cell in matrix or vector contains same probability.
'input' Function takes initial parameters from InitParams argument.
[Required with InitMethods ('input')] When InitMethods specifies 'input', this argument specifies the initial parameters for the models:
InitParams Item Description
init_state_probability_vector Vector that contains initial state probabilities for model.
state_transition_probability_matrix Matrix that contains state transition probabilities for model.
observation_emission_probability_matrix Matrix that contains observation emission probabilities for model.
For example, if the NumberHiddenStates argument specifies three hidden states and two observed symbols ('yes' and 'no'), these are the InitParams values:
InitParams Item Value
init_state_probability_vector '0.3333333333 0.3333333333 0.3333333333'
state_transition_probability_matrix '0.3333333333 0.3333333333 0.3333333333;

0.3333333333 0.3333333333 0.3333333333;

0.3333333333 0.3333333333 0.3333333333'

observation_emission_probability_matrix 'no:0.25 yes:0.75; no:0.35 yes:0.65; no:0.45 yes:0.55'

The sum of the probabilities in each row for the initial state probabilities, state transition probabilities, or observation emission probabilities parameters must round to 1.0. The number of states and the number of observed symbols must be consistent with the NumberHiddenStates argument and the observed symbols in the input table; otherwise, the function displays error messages.