Selective Inference

Selective inference (SI) is a statistical framework addressing the challenges of making valid inferences after data-driven model selection or hypothesis generation. Current research focuses on applying SI to diverse areas, including anomaly detection, time series analysis, and the evaluation of AI models (e.g., using recurrent neural networks and diffusion models), often incorporating techniques like post-processing debiasing or calibration regularization to improve reliability. This work is significant because it provides rigorous statistical tests for results obtained through complex data analysis pipelines and AI-driven hypothesis generation, improving the trustworthiness and interpretability of findings across various scientific disciplines and applications.