Structure of Nonlinear Node Embeddings in Stochastic Block Models

This repository provides the source code to reproduce the experimental results contained in "Structure of Nonlinear Node Embeddings in Stochastic Block Models."

Requirements

The necessary requirements can be found requirments.txt. We recommend using virtualenv to create a virtual environment, install the necessary requirements, and install the provided deepwalk package.

pip install -r requirements.txt 

In the project directory containing the setup.py file, run

pip install .

to install the deepwalk package.

Description of Files

• deepwalk : contains code to run the DeepWalk and SGNS models. It also contains code to generate a co-occurrence matrix from a graph. Usage can be found in the class and function docstrings.

  • The models sub-package contains code for the deepwalk and negative_sampling classes.
  • The walks sub-package contains code to generate a co-occurrence matrix.

• notebooks : Contains Jupyter notebooks that reproduce the figures and experimental results in the papaer

  • "Brute Force Optimization" and "Brute Force Optimization - Unequal sized clusters" reproduces the brute force optimization results the plots in Figure 1 in the paper.
  • "Sensitivity to Number of Negative Samples" reproduces the plots shown in Figure 2 of the paper.