Databricks Feature stores

[Ariane]

After exploring the feature store and how it works I have some concerns

1. With each data refresh, there are possibilities for a change in feature values. Does Databricks feature store allow to alter the feature table in case the feature values have changed? or do you have to create another one?

2. In case there is new data for a new or existing client, can I override the existing feature store or can the feature store allow adding new features or altering existing ones?

[Kumaran]

Hi @Ariane,

Thank you for posting your question in Databricks community.

Regarding your concerns:

1. Yes, Databricks feature store allows you to update the feature table in case the feature values have changed. You can update the data in the Delta table underlying the feature table, then update the feature table accordingly with the new data.

2. Yes, Databricks feature store allows you to add new features or alter existing ones. When new data is available, you can create a new version of the feature table with the updated data and metadata. You can also include new features or alter existing ones by updating the schema of the feature table and triggering a new computation on the updated data.

[Ariane]

Can you please point me to a documentation or a demo or anything else that shows how the feature stores are altered? so that I can see how it is done

[Kumaran]

Hello @Ariane,

Could you check the same by downloading the ebook : The Comprehensive Guide to Feature Stores here ?