Awesome
Reliable Predictive Inference
An important factor to guarantee a responsible use of data-driven recommendation systems is that we should be able to communicate their uncertainty to decision makers. This can be accomplished by constructing prediction intervals, which provide an intuitive measure of the limits of predictive performance.
This package contains a Python implementation of Orthogonal quantile regression (OQR) [1] methodology for constructing distribution-free prediction intervals.
Orthogonal Quantile Regression [1]
OQR is a method that improves the conditional validity of standard quantile regression methods.
[1] Shai Feldman, Stephen Bates, Yaniv Romano, “Improving Conditional Coverage via Orthogonal Quantile Regression.” 2021.
Getting Started
This package is self-contained and implemented in python.
Part of the code is a taken from the calibrated-quantile-uq package available at https://github.com/YoungseogChung/calibrated-quantile-uq.
Prerequisites
- python
- numpy
- scipy
- scikit-learn
- pytorch
- pandas
Installing
The development version is available here on github:
git clone https://github.com/shai128/oqr.git
Usage
OQR
Please refer to oqr_synthetic_data_example.ipynb for basic usage. Comparisons to competitive methods and can be found in display_results.ipynb.
Reproducible Research
The code available under /reproducible_experiments/ in the repository replicates the experimental results in [1].
Publicly Available Datasets
-
Blog: BlogFeedback data set.
-
Bio: Physicochemical properties of protein tertiary structure data set.
-
Kin8nm: A variant of Kin family of datasets.
-
Naval: Condition based maintenance of naval propulsion plants data set.
-
Facebook Variant 1 and Variant 2: Facebook comment volume data set.
Data subject to copyright/usage rules
The Medical Expenditure Panel Survey (MPES) data can be downloaded using the code in the folder /get_meps_data/ under this repository. It is based on this explanation (code provided by IBM's AIF360).