Interpretable models for extrapolation in scientific machine learning

Abstract

On average, simple linear models perform equivalently to black box machine learning models on extrapolation tasks.