Granger Busca

Code for our paper Fast Estimation of Causal Interactions using Wold Processes to appear at NIPS2018.

Installation Instructions

Dependencies

• Python version 3.5 or later
• C++

Python dependencies:

• Cython (0.26.1)
• matplotlib (2.1.0)
• numpy (1.13.3)
• pandas (0.21.0)
• scipy (1.0.0)
• tables (3.4.2)
• tick (0.3.0.0)

(Also available in requirements.txt)

Compiling

To compile, run:

$ make

As a python module

To install as a module execute the following:

$ python setup.py install

Examples

Simple example on how to fit Granger-Busca on a dataset available in the file (requires jupyter notebook):

• Example Granger-Busca.ipynb

Other notebooks:

• Testing with data.ipynb -- Shows the model executing on a small subsample of memetracker data
• View data.ipynb -- Explores the memetracker results

Snap Baselines

The scripts used to extract results from the snap datasets are available in the snap-baselines/ folder.