The Feast Feature Store and Snowflake

The Providence Athenaeum Library

A feature is a property of a data product that has been reduced into a form to be used for Machine Learning. The feature store contains values transformed via standard methods, such as providing boolean 0 or 1 values for specific properties such as is_blue or is_read or enumerating values like 0 for blue or 1 for red. Additionally, serial numbers can be turned into features by encoding the z-value or creating a mean [1].

Snowflake integrates with the open-source feature store technology Feast [2, 3].

References

1. Google IO (2023) The Transform TFX Pipeline Component

2. Feast (2023) Open Source Feature Store for Production ML

3. M. Adkins (2022) Building machine learning features in Snowflake with open-source feature store Feast