[
https://issues.apache.org/jira/browse/ATLAS-3432?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16946268#comment-16946268
]
Karthik Manamcheri commented on ATLAS-3432:
-------------------------------------------
I think the type system should comply with the standard proposed in [this blog
post.|https://blog.cloudera.com/creating-an-open-standard-machine-learning-governance-using-apache-atlas/]
If not, I would like to understand why we choose not to follow that system and
if there's a particular case.
The reason I am pushing for the standard is that it might get difficult to
update the type system in the future for existing components. I'd rather start
small and add new types than have to remove types. The code-review in the form
today has "ML Model" as a type (which is not present in the blog post
proposal). Its easier to add a new type than to remove a type. That's my $0.02.
Thoughts [~sarath] [~amestry] [~linaataustin]?
> Add Machine Learning Governance integration
> -------------------------------------------
>
> Key: ATLAS-3432
> URL: https://issues.apache.org/jira/browse/ATLAS-3432
> Project: Atlas
> Issue Type: New Feature
> Components: atlas-core
> Affects Versions: 3.0.0
> Reporter: Na Li
> Assignee: Na Li
> Priority: Major
>
> Managing machine learning models and its deployments are getting more
> important over time. Adding their metadata into Atlas is the key step.
> This update adds new entity types for ML governance, and implements the Hook
> defined for ML governance.
> * Atlas entities for machine learning metadata are defined.
> * The Hook API is defined for customer governance service to call to store ML
> metadata.
> * An implementation of the Hook API is provide that converts machine learning
> metadata to lineage notifications and send them to Atlas.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
