Skip to content

RevIN

Home / Concepts / RevIN

RevIN

Auto-generated stub. Edit this file to add more details.

A reversible instance normalization technique designed to handle non-stationarity and scale variation in time series data.

Why It Matters

The study provides empirical validation that RevIN, originally for traditional time-series models, is highly effective for improving zero-shot generalization in Foundation Models.

Evidence

we empirically establish RevIN as the most efficient approach, reducing zero-shot MASE by 89% relative to raw baseline

Metadata & Links