SLCP

Code for the paper Stable Localized Conformal Prediction via Transduction.

This repository contains the core implementation of SLCP, together with the simulation and real-data experiment scripts used in the paper.

Structure

• Main/: main method implementation and supporting modules.
• SimuAnalysis/: simulation experiments.
• RealAnalysis/: real-data experiments.
• Dataset/: datasets used by the experiment scripts.

Dependencies

This repository does not currently include a requirements.txt. Based on the code, the main dependencies include:

• numpy
• pandas
• scipy
• scikit-learn
• matplotlib
• torch

Some scripts may also require additional dataset-specific support, for example pyreadstat for reading .sav files via pandas.

Example Usage

Run one simulation setting with:

python SimuAnalysis/run_shared.py quad 30 500

Run real-data experiments with scripts such as:

python RealAnalysis/Crime.py
python RealAnalysis/Protein.py
python RealAnalysis/Achieve.py
python RealAnalysis/Derma.py
python RealAnalysis/Tissue.py

Notes

• Some image experiments use pretrained checkpoints stored in RealAnalysis/Para/.
• Experiment outputs may be written to directories such as SimResult/, Log/, and Figure/.