This code is a supplementary material for the paper "Ellipsoidal conformal inference for Multi-Target Regression" accepted in COPA 2022. It includes necessary source code for reproducing synthetic data results, as well as 3 real data set results (other data sets were omitted for size restrictions).
This project is composed of 3 folders:
- code : contains main code for synthetic data and real data sets' experiments.
- utilities : contains files with essential functions used in the "code" folder for simulating synthetic data, real data preprocessing, empirical and ellipsoidal non-conformity measures' functions.
- input : contains a folder for each real data set with its data and config files, and stores log and visualization files when produced by the "code" folder. There are 3 data sets to test with (residential building, enb and scpf).
To reproduce the results in the paper, you can run the files in the "code" folder :
- conformal_synthetic_data : produces results for synthetic data. Visualization files are stored in the "code" folder.
- conformal_real_data : produces results for real data sets in the "input" folder. You can run the "run.sh" or "run.bat" file to execute the code.
pandas==0.24.2 matplotlib==3.0.3 tensorflow==2.0.1 scipy==1.2.1 copulae==0.6.0 numpy==1.16.2 scikit_learn==1.0.2
- Code related to data preprocessing and empirical copula non-conformity measures is taken from our previous work.
- Results may slightly vary from presented results in the paper.