Statistically Calibrated Decision Risk Optimization
By integrating confidence modeling, evidence combination, and knowledge incorporation, this direction provides robust and interpretable feature representations for handling heterogeneous and low-quality medical data, significantly improving the accuracy and safety of downstream tasks.
Conformal Prediction
Focuses on statistically valid prediction sets with guaranteed error control, enabling reliable and risk-aware decision-making.
Multiple Hypothesis Testing
Focuses on controlling false discovery rates in large-scale decision scenarios, supporting safe and statistically principled clinical inference.
