Brief

A new study evaluating machine learning models for chronic kidney disease (CKD) risk prediction found that models achieving near-perfect performance on internal test sets failed to generalize to external data. The research highlighted significant drops in accuracy and calibration when models were applied to a different patient cohort, revealing a critical gap in deployment readiness. The authors emphasize the necessity of evaluating calibration stability and uncertainty quantification on external datasets before any clinical prediction model is considered for deployment. AI

Researchers have developed a new active learning framework called Cumulative Active Meta-Learning (CAML) to improve the robustness of machine learning models against spurious correlations. CAML treats each active learning round as a meta-learning task, using queried samples to refine the model's inductive bias rather than just updating its likelihood. This cumulative approach captures sequential dependencies between learning rounds, leading to significant accuracy improvements for minority groups on various benchmarks. AI