Mentored Projects

This page maintains a list of mentored project ideas that contributors can work on if they are interested in contributing to the NetworkX project. Feel free to suggest any other idea if you are interested on the NetworkX GitHub discussions page

These ideas can be used as projects for Google Summer of Code, Outreachy, NumFOCUS Small Development Grants and university course/project credits (if your university allows contribution to open source for credit).

Community Detection Algorithms

  • Abstract: Community detection involves a set of algorithms in network science which deal with grouping nodes from a network according to their similar properties such as belonging to dense clusters. NetworkX already contains a variety of community detection algorithms dealing with computing the community structure of a network. There are also multiple PRs/issues which deal with adding the Louvain community detection algorithm to NetworkX, e.g. #1090, #1092 #951. Users who want to work with NetworkX and Louvain Community Detection often use https://github.com/taynaud/python-louvain. This project would focus on getting Louvain community detection algorithms implemented into NetworkX.

  • Recommended Skills: Python, graph algorithms

  • Expected Outcome: We would like to see Louvain community detection implemented inside NetworkX, or construct code and documented examples in NetworkX that would interface with other Louvain projects.

  • Complexity: Moderate

  • Interested Mentors: @dschult, @MridulS,

Pedagogical Interactive Notebooks for Algorithms Implemented in NetworkX

  • Abstract: NetworkX has a wide variety of algorithms implemented. Even though the algorithms are well documented, explanations of the ideas behind the algorithms are often missing and we would like to collect these, write Jupyter notebooks to elucidate these ideas and explore the algorithms experimentally, and publish the notebooks at https://github.com/networkx/notebooks. The goal is to gives readers a deeper outlook behind standard network science and graph theory algorithms and encourage them to delve further into the topic.

  • Recommended Skills: Python, Jupyter notebooks, graph algorithms.

  • Expected Outcome: A collection of Interactive Jupyter notebooks which explain and explore network algorithms to readers and users of NetworkX. For example, see this notebook on Random Geometric Graphs

  • Complexity: Depending on the algorithms you are interested to work on.

  • Interested Mentors: @dschult, @MridulS, @rossbar

Directed Version of Traveling Salesman Problem

  • Abstract: NetworkX has recently added a couple methods for solving the Traveling Salesman Problem (see #4607). The best approximation for undirected graphs is the Christofides method. But the best algorithm for directed graphs is by Asapour et.al. and has not yet been implemented. The goal of this project is to learn the API used for implemented methods and then implement the Asadpour method for directed graphs with similar API. Other even more recent papers discussing algorithm improvements for directed TSP (also called Asymmetric TSP or ATSP) include Svensson and Traub. The Traub paper may be most useful for implementing the algorithm as all three are focused on proving asymptotic computation requirements rather than coding.

  • Recommended Skills: Python, graph algorithms

  • Expected Outcome: A new function in NetworkX which implements the Asapour algorithm.

  • Complexity: Moderate

  • Interested Mentors: @dschult, @MridulS, @boothby,

Implement the VF2++ Graph Isomorphism Algorithm

  • Abstract: The Graph Isomorphism Problem is a famous difficult network problem at the boundary between P and NP-Complete. The VF2 algorithm is included with NetworkX in a recursive formulation. There is an improved version of this algorithm called VF2++ which we intend to implement. We have early attempts at a nonrecursive version of the main algorithm that also address subgraph isomorphism and subgraph monomorphism. This project involves fully implementing them and extending to directed and multigraph settings.

  • Recommended Skills: Python, graph algorithms

  • Expected Outcome: A new set of functions in NetworkX that implement the VF2++ algorithm for all problem and graph types in a nonrecursive manner.

  • Complexity: Moderate

  • Interested Mentors: @dschult, @MridulS, @boothby,

Project Idea Template

  • Abstract:

  • Recommended Skills:

  • Expected Outcome:

  • Complexity;

  • Interested Mentors: