GraphBLAS is a powerful toolkit for building out graph algorithms. We’ll build out a couple commonly used graph algorithms including some centrality estimating methods, and a more complex algorithm such as PageRank. Finally, we’ll work on a problem which harnesses either iterative solvers or the direct solvers from SuiteSparse.
Ask questions during the workshop: https://pigeonhole.at/JULIA9
Make sure to register for JuliaCon to get access to all of the resources: https://juliacon.org/2022/tickets/
SuiteSparseGraphBLAS.jl and the surrounding Julia GraphBLAS ecosystem allow Julia users to succinctly express graph and sparse matrix algorithms without sacrificing speed. In this workshop we will introduce the basic usage, terminology, and features of the package by building several graph and sparse matrix algorithms.
These graph algorithms take advantage of graph matrix duality to express basic graph operations like breadth first search as a matrix-vector multiplication under a particular semiring. We will start with a simple breadth first search implementation before moving on to a few increasingly complex functions including a centrality measure (triangle centrality) and one or more larger algorithms like PageRank or a message passing neural network.
The last problem we will workshop involves using either iterative solvers like Krylov.jl or the direct solvers from SuiteSparse.jl, showcasing the unique ability to use these solvers on GraphBLAS matrices.