The module outlier_detector.py has two classes: (1) Detector and (2) ScoreModel. The Detector class has multiple functions to perform noise detection. The entry point function to use this class is “purify”
All benchmark tests and used dataset can be found in “outlier_detector/tests” To run test problems run the scripts listed below. These scripts conduct multiple numerical experiments and save results in a folder named “results”. The datasets include 2 synthetic problems (1D and 6D datasets), and a real-world problem (public supply in the southwest USA).
- Run one_dimension_case.py
- Run hartmann_6d.py
- run evaluating_data_model.py
- fig_effect_of_seed_number.py
- fig_effect_of_noise_signal_ratio.py
- run fig_compare_smapler_functions.py
A complete copy of the dataset for the Public Supply can be found at https://doi.org/10.5066/P9FUL880. A subset of the data for California, Arizona, and Nevda are extracted and used for the testing (south_westh.csv)
- Run ca_case.py
- Run evaluate_equifinality.py
- run long_run.py