Examples

Practical examples demonstrating different conformal anomaly detection approaches.

Getting Started

Example	Difficulty	What You'll Learn
Classical Conformal	Beginner	Basic split conformal detection, FDR control, p-value interpretation
Data-Efficient Resampling	Intermediate	CV+, Jackknife+, and JaB+ for small-data calibration

Advanced Strategies

Example	Difficulty	What You'll Learn
Conditional Conformal	Intermediate	Conditionally calibrated conformal p-values with BH-style FDR selection

Special Topics

Example	Difficulty	What You'll Learn
Weighted Conformal	Advanced	Handling distribution shift between training and test data
FDR Control	Intermediate	Multiple testing correction, Benjamini-Hochberg procedure

What Each Example Covers

Classical Conformal - Start here if you're new to nonconform. Learn the core workflow: wrap a detector, compute p-values, and apply FDR control. Includes visualization of results.

Data-Efficient Resampling - Learn CV+, Jackknife+, and JaB+ together as one family of strategies for using scarce training data efficiently. Start here when a fixed calibration holdout would cost too much power.

Conditional Conformal - Apply conditional calibration maps (mc, simes, dkwm, asymptotic) to empirical conformal p-values, then run BH-style FDR selection.

Weighted Conformal - Handle covariate shift scenarios where your test data comes from a different distribution than your training data. Essential for real-world deployment.

FDR Control - Deep dive into False Discovery Rate control. Understand when to use BH vs weighted methods and how to evaluate FDR performance.

Prerequisites

All examples assume you have installed nonconform with the PyOD and data extras:

pip install "nonconform[pyod,data]"

Most examples use the PyOD LOF detector and benchmark datasets from oddball. Each example is self-contained and can be run independently.