PAPER

Implementation of Gibbs sampler for computing posterior root probabilities under the PAPER (preferential attachment plus Erdos--Renyi) model for random networks.

See details in the arXiv paper: https://arxiv.org/abs/2107.00153

Documentation

Installation

$ pip install PAPER

Usage

>>> from PAPER.gibbsSampling import gibbsToConv
>>> from PAPER.tree_tools import createNoisyGraph
>>> graf = createNoisyGraph(n=100, m=200, alpha=0, beta=1, K=1)[0]
>>> mcmc_out = gibbsSampling.gibbsToConv(graf, DP=False, method="full",
                   K=1, tol=0.1)

See example.py for interpreting the inference output. Some sample network datasets are provided.

Notes

• No preprocessing required on input graph. If the input graph is disconnected, the largest connected component is used.
• The algorithm performs roughly 1 outer Gibbs iteration in 1 second on a graph with 10,000 edges. The number of iterations to convergence depends on the input graph.