1.0 - 8.00 - LDA Output - Teradata Vantage

Teradata® Vantage Machine Learning Engine Analytic Function Reference

Product
Teradata Vantage
Release Number
1.0
8.00
Release Date
May 2019
Content Type
Programming Reference
Publication ID
B700-4003-098K
Language
English (United States)

Output Message Schema

Column Data Type Description
message VARCHAR Reports iteration steps and perplexity of model.

Perplexity formula:

perplexity = 2 H (p) = 2 x p (x) log2 p (x)

where H (p) is the entropy of the distribution.

Although perplexity varies with training documents, you can use perplexity to find the best model for a specified set of training documents: Create models for several subsets of the training documents and then choose the model with the lowest perplexity.

ModelTable Schema

Column Data Type Description
topicid INTEGER Internally created topic identifier.
value_col VARBYTE(n) Model in binary format, which is not readable. To see binary contents, use LDATopicSummary function.
Column size depends on InputTable size and may be very large. If any cell of this column exceeds 64 KB, you cannot use ModelTable with LDAInference or LDATopicSummary function.

OutputTable Schema

This table appears only with the OutputTable argument.

Column Data Type Description
docid Same as doc_id_column in input table Document identifier from input table.
topicid INTEGER Topic identifier from ModelTable.
topicweight DOUBLE PRECISION [Column appears number of times specified by OutputTopicNum argument.] Topic weight.
topicwords VARCHAR [Column appears number of times specified by OutputTopicWordNum argument.] Topic words in document, separated by commas.