Prediction of microstructural representativity from a single image

This master thesis conducts a qualitative content analysis of 71 recent papers from the NeurIPS “Datasets and Benchmarks” track to investigate the operational meanings, or “notions”, that researchers assign to representativity, similarity and diversity

This simulation shows how objective reference baselines can inform safe AI deployment in new target populations, and why representativity must be defined relative to estimands, audit designs, estimators, uncertainty procedures, and validation claims.