Research
You can also find my articles on my Google Scholar profile.
Current projects
- Inference based on confidence sets for distributions
- Developing methods for deriving confidence intervals for various functionals (e.g., mean or median) under user-specified assumptions (e.g., finite variance or tail behavior), using confidence sets for CDFs.
- This offers a flexible inference strategy that reduces dependence on strict assumptions and enhances applicability across diverse contexts.
- Application of the post-selection conformal method: police officer safety data
- Police officers in the USA face significant risks during service calls, making accurate risk assessments crucial for their safety.
- Black-box machine learning models improve risk assessment by identifying patterns in call data.
- Conformal inference procedures provide prediction sets with target coverage probability guarantees but traditionally require fixed coverage levels.
- The aim is to apply a conformal inference procedure with a post-selection guarantee for coverage levels (developed here), allowing adjustable coverage levels while maintaining the usual guarantees.
- This approach enhances risk forecasts, improving officer safety during service calls.
- Random forests and decision trees
- Focusing on density estimation trees (DETs), a data-driven partitioning approach for tree-structured density estimation.
- Establishing the necessary conditions for DETs to be consistent estimators, under different metrics such as \(L^2\) norm or KL divergence.
- Exploring methods to quantify uncertainty in density estimates from DETs, enhancing their reliability and interpretability in machine learning applications.