Source code for confidence set analysis on the Diffusion Source Identification (DSI) problem. Paper presented at ICML 2021 can be found at ICML 2021's Paper List or on arXiv.
This is an updated version of the code. Previous version can be found in the legacy directory.
The code can be installed by running the following commands
git clone https://github.com/lab-sigma/Diffusion-Source-Identification
cd Diffusion-Source-Identification
pip install .
The package is named ``diffusion_source''. The relevant subpackages are summarized below.
Provides a wrapper around a networkx graph used by the infection model. A superset of the synthetic networks used in the paper are provided, as well as classes for importing a graph from adjacency formats and an already created networkx graph.
A base abstract infection model class is provided as ``InfectionModelBase'' that includes a definition of the confidence set function for arbitrary diffusion processes.
A number of discrepancy functions are written for use when running the confidence set algorithm
A set of tools for displaying example infected sets and confidence sets, as well as generating summary figures for large scale model tests.