epfl-lasa / Changepoint Detection

Bayesian Online Multivariate Changepoint Detection Algorithm

Student: Ilaria Lauzana
Supervisors: Nadia Figueroa, Jose Medina

This repository contains the implementation of the Bayesian Online Multivariate Changepoint Detection algorithm, proposed by Ilaria Lauzana, Nadia Figueroa and Jose Medina.

We provide 3 implementations:

• matlab
• python
• ros node to detect changepoints from streaming data (online_changepoint_detector)

You can find each implementation in its corresponding folder:

Structure

.
├── README.md
└── matlab
    ├── README.md
    │   └── code
    │   └── lightspeed
└── python
    ├── python-univariate
        ├── README.md
        │   └── bayesian_changepoint_detection
    ├── python-multivariate
└── online_changepoint_detector
    ├── CMakeLists.txt
    ├── package.xml
    └── scripts
└── data
└── results - figures
└── report-project-changepoint
    ├── README.md
    ├── main.tex
    └── references

Instructions:

Matlab

The matlab implementation is a self-contained code, no dependencies are needed. Except for the lightspeed toolbox, which is provided within the folder.

In order to run the changepoint detector, run the follwing script found in ./matlab/code/:

> gaussdemo_multi.m

Python

For the python implementation, install the following python libraries for linear algebra, machine learning methods and plotting:

$ sudo apt-get install python-numpy python-scipy python-matplotlib ipython ipython-notebook python-pandas python-sympy python-nose

Then install seaborn:

$ sudo pip install seaborn

Once installed, you can test the following example, found in ./python/python-multivariate/:

$ python ./example_stream_data.py

If something is not working, try updating numpy, this generallt fixes the problem:

$ sudo pip install numpy --upgrade

Ros Node

Follow the README file in ./online-changepoint-detector/, must have all dependencies installed for the python implementation.

... piece of 🍰