Skip to content

Ability to update recommenders incrementally #5318

Description

@reckart

Is your feature request related to a problem? Please describe.
Currently, all recommenders train totally from scratch when annotations change. However, in some cases, it might be possible to just update the existing model.

Describe the solution you'd like
Implement the necessary APIs for recommenders to generate training instances only for that data that has changed and update themselves using these.

Metadata

Metadata

Assignees

Projects

Status
🏃‍♀️ In progress

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions