adaptive-conformal-inference 0.1.0

Adaptive Conformal Inference Under Distribution Shift

adaptive-conformal-inference

Python implementation of Adaptive Conformal Inference (ACI) from:

Isaac Gibbs and Emmanuel J. Candès. Adaptive Conformal Inference Under Distribution Shift. NeurIPS 2021. arXiv:2106.00170

ACI adapts conformal prediction sets to distribution shift by tracking a single parameter that adjusts the quantile level online, provably achieving target coverage over long time intervals.

Installation

pip install adaptive-conformal-inference

For running the example scripts (stock data + plots):

pip install adaptive-conformal-inference[examples]

Quick Start

from aci import ACITracker

# Core: model-agnostic alpha tracker
tracker = ACITracker(alpha=0.1, gamma=0.005)

for t in range(T):
    # Your conformal prediction logic here...
    err_t = 1.0 if y_true not in prediction_set else 0.0
    tracker.update(err_t)
    next_alpha = tracker.alpha_t  # use this for the next prediction set

Modules

Module	Description
aci.tracker.ACITracker	Core alpha update (Simple and Momentum rules)

Non-essential forecasting/data pipelines are intentionally kept in examples/:

• examples/figure1/ (self-contained Figure 1 example files)
• examples/figure2/ (self-contained Figure 2 example files)

Reproducing Paper Figures

# Figure 1: Volatility prediction with normalized score (4 stocks)
python examples/figure1/reproduce.py

# Figure 2: Volatility prediction with unnormalized score (4 stocks)
python examples/figure2/reproduce.py

Reproduced Figures From the Paper

Figure 1 (Normalized Score)

Our reproduction	Paper figure

Figure 2 (Unnormalized Score)

Our reproduction	Paper figure

Citation

  @article{gibbs2021adaptive,
    title={Adaptive conformal inference under distribution shift},
    author={Gibbs, Isaac and Candes, Emmanuel},
    journal={Advances in Neural Information Processing Systems},
    volume={34},
    pages={1660--1672},
    year={2021}
  }