ralf: Real-time, Accuracy Aware Feature Store Maintenance
Sarah, PhD will introduce the notion of feature store regret that helps evaluate feature quality of different maintenance policies.
Sarah, PhD will introduce the notion of feature store regret that helps evaluate feature quality of different maintenance policies.
Feature stores are becoming ubiquitous in real-time model serving systems, however there has been limited work in understanding how features should be maintained over changing data. In this talk, we present ongoing research at the RISELab on streaming feature maintenance that optimizes both resource costs and downstream model accuracy. We introduce a notion of feature store regret to evaluate feature quality of different maintenance policies, and test various policies on real-world time-series data.
