Prediction-Powered Inference
This enables researchers across fields like proteomics and ecology to draw valid, data-efficient conclusions using ML, representing a novel methodological advancement rather than an incremental improvement.
The paper tackles the problem of performing valid statistical inference when experimental data is supplemented with machine-learning predictions, resulting in a framework that provides provably valid confidence intervals for various statistical quantities without assumptions on the ML algorithm, with more accurate predictions leading to smaller intervals.
Prediction-powered inference is a framework for performing valid statistical inference when an experimental dataset is supplemented with predictions from a machine-learning system. The framework yields simple algorithms for computing provably valid confidence intervals for quantities such as means, quantiles, and linear and logistic regression coefficients, without making any assumptions on the machine-learning algorithm that supplies the predictions. Furthermore, more accurate predictions translate to smaller confidence intervals. Prediction-powered inference could enable researchers to draw valid and more data-efficient conclusions using machine learning. The benefits of prediction-powered inference are demonstrated with datasets from proteomics, astronomy, genomics, remote sensing, census analysis, and ecology.