distribution-free

Wrapper for a PyTorch classifier which allows it to output prediction sets. The sets are theoretically guaranteed to contain the true class with high probability (via conformal prediction).

Conformalized matrix completion

Guide and Guard: Prior-Guided Virtual-Contrast MRI Synthesis with Conformal Flags for Missed Enhancement (Under review at IEEE ISBI 2027)

An Independence Test based on Data-Driven Tree-Structured Representations.

Example on task completion times stressing the importance of distribution-free intervals, as well as .py file containing code to calculate such intervals.

Self-Organized Conformal Prediction (NeurIPS 2026): local conformal calibration on groups discovered by a self-organizing map.

Guide and Guard: Prior-Guided Virtual-Contrast MRI Synthesis with Conformal Flags for Missed Enhancement

Distribution-free miss guarantees (conformal risk control) for object detectors in medical imaging: thresholds with a guaranteed miss rate, their price in false detections, desktop app for Windows.

Distribution-free test for general differences in two populations

Conformal prediction from scratch: the exact finite-sample coverage law, the 2012 deployment that broke it (Hurricane Sandy included), and the online repair

A method for cold-deck imputation of a continuous distribution from binned incomes, using a real-world reference data set

ConformalLab — Interactive Conformal Prediction & Distribution-Free Inference Laboratory. 6 modules: split conformal, ACI, classification sets, coverage guarantees, CQR, Mondrian CP. Zero dependencies.

A distribution-free conformal prediction certification layer for safety-critical ordinal classification. Guarantees contiguous prediction sets, marginal/group-conditional validity, true-label coverage under banded reporting noise, and severity-weighted risk control (fatal-omission bounds) with an explicit additive slack budget.

Topology-invariant spatiotemporal forecasting with distribution-free, coverage-guaranteed prediction intervals at unseen sensor locations under dynamic (add/remove) topology. Reference code + experiments for the OmniST paper.

ConformalForge v0.2.0 — 分布无关不确定性量化（分类/回归融合旗舰，作者 晨星）