ML - Can I import and run a model that uses functionality not native in Vertica?
emichaud
Vertica Employee Employee
If I'm importing a model, let's say from Jupyter, and the script in Python or R contains some library that has some functionality or in fact some algorithm that Vertica does not contain, how we will be able to run it? If we want to make the deployment?
Asked during the BDC session Model Management and Data Preparation
@WD @Arash_Fard
0
Answers
We have started to add support for PMML standard in Vertica 10.0.
"PMML provides a way for analytic applications to describe and exchange predictive models produced by data mining and machine learning algorithms." (Ref: https://en.wikipedia.org/wiki/Predictive_Model_Markup_Language)
As long as a trained ML model can be converted to its PMML equivalent, regardless of its training platform and libraries used for its training, it should be possible to deploy it into any other PMML supporting platform for scoring. There are open source libraries both in R and Python for converting their models into PMML. Therefore in long term, the users would be able to convert their external models to PMML and then import them into Vertica for in-database scoring.