Skip to content

Building a Faster NetworkX-Like Library Using GraphBLAS

Short Talk at 2:57PM EDT

This is a call for engagement to help with graphblas-algorithms!
We'll also have weird math and weird Python syntax,
so you can also come for the fun of getting your mind blown :)

GraphBLAS is a mathematical abstraction layer designed to express many
graph algorithms in the language of linear algebra. It's weird and written in C.
It is also incredibly fast and powerful.
Graph analytics has applications across industries and sciences--more than you might expect--and we want to get the power of GraphBLAS into the hands of Pythonistas and practitioners.

Our package, python-graphblas, allows us to write GraphBLAS in the style of math, not the C API.
Counterintuitively, by introducing some unconventional notation, we were able to make our library more Pythonic.
However, python-graphblas is still low-level, and most users probably won't use it directly.
Instead, most people will probably want to use graphblas-algorithms,
a library that looks like NetworkX, the most popular graph library in Python.

graphblas-algorithms is young.
It is growing and has great potential, but it could use more help.
Who can help? Anybody who is interested! We have a well-defined target--the NetworkX API and its tests--so
it's very easy to find things to do.
We actually find it fun like solving puzzles, and there is still a lot of low-lying fruit.
If you don't know much about graphs or how to contribute to open-source, this could be a great way to learn!

Presented by

Erik Welch