This repository includes all the material used in the experimental study of the work on Covariance Change Point Detection (CPD) for Graph Signals. It also provides additional content related to CPD for Graph Signals and Graph Stationarity.
The repository is organized as follows:
Detailed list
- utils.py: file writing and processing.
- signal_related.py: signal generation and modification.
- graph_related.py: graph generation and modification.
- result_related.py: metrics computation, processing and storage.
- custom_cost_functions.py: cost function classes, from the ruptures BaseCost class.
- numba_cost_functions.py: numba-compatible implementations of the cost functions.
- running_cpd.py: utils and dynamic programming implementation for CPD solving.
- rpy2_related.py: utils and wrapping functions for the apllication of the Graph Lasso algorithm from [Friedman2008] and the covcp method from [Avanesov2018].
Detailed list
- run_cpd_cmd_line.py: produces predictions files for the target signals and methods.
- compute_metrics_cmd_line.py: produces metrics files based on prediction files.
Detailed list
- covariance_matrix_cpd.ipynb: contains the main utilities such as predictions generation (same as but more flexible than run_cpd_cmd_line.py), metrics computation (same as but more flexible than compute_metrics_cmd_line.py), plotting utils and cells as well as experiments presentation and visualization.
- cpd_playground.ipynb: additional content related to CPD (search algorithms comparions) and the application to local mean change for graph signals, through the Graph Fourier Scan Statistic [Ferrari2019]
- graph_stationarity.ipynb: additional content related to graph stationarity. Includes the appendix material of the work on Covariance CPD.
- real_dataset_preprocessing: utils, visualization and preprocessing for different CPD real datasets.
🚧 In progress 🚧
[Avanesov2018] V. Avanesov and N. Buzun, Change-point detection in high-dimensional covariance structure, Electronic Journal of Statistics, vol. 12, no. 2, pp. 3254–3294, Jan. 2018, Publisher: Institute of Mathematical
[Ferrari2019] A. Ferrari, C. Richard, and L. Verduci. Distributed Change Detection in Streaming Graph Signals. In 2019 IEEE 8th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), pages 166–170,
[Friedman2008] J. Friedman, T. Hastie, and R. Tibshirani, Sparse inverse covariance estimation with the graphical lasso, Biostatistics, vol. 9, no. 3, pp. 432–441, July 2008.