News related to the GraphBLAS API and its community. GraphBLAS implementations, graph algorithms in linear algebra and graph processing on specialized hardware.
Here's a link to my JuliaCon 2023 keynote talk on @GraphBLAS and its impact on @JuliaLanguage,
@MATLAB, https://t.co/tCZhkSiKPx (and thus
@LeidosInc), @falkordb (formerly RedisGraph) and more.
https://t.co/NRDJ45Z3BQ
(reposted with link fixed)
Speeding up graph algorithms in python by up to 21,000x. 🔥
via a NetworkX+@anacondoninc+@GraphBLAS collaboration: NetworkX (@networkx_team) performance boost using the python interface to SuiteSparse:@GraphBLAS. @pydata talk by @eriknwelch & Jim Kitchen https://t.co/dE8Q5ngBnc
I'm at @Supercomputing 2022 in Dallas and would be happy to chat about @LDBCouncil benchmarks and @GraphBLAS applications. Send me a DM if you're interested.
@NicoleHemsoth I'll be at the Graph Toolkit and @GraphBLAS BoF: https://t.co/NG1Mv8I3km
GraphBLAS uses sparse linear algebra over semirings to create graph algorithms, and appears in @Redis, Anaconda, @MATLAB, and other places. Should be fun!
@GraphBLAS SIAM News article: graph algorithms via linear algebra over semirings, using sparse matrices. Already a part of @MATLAB, @Redis, @Anaconda conda-forge, https://t.co/TcSDBjVfz8 (AI on graphs), @JuliaLanguage, ... https://t.co/Q0ycxALYiV
I have 2 PhDs who will finish this year: @aznaveh1 working on parallel sparse LU & @Jin_hao_Chen working on exact update/downdate of sparse LU and Cholesky. Both have worked with @GraphBLAS too. Great work! Both looking for positions (academic & non-academic) DM for details
@GraphBLAS speeds up sparse matrices in Julia by up to 1000x (for C[I,J]=A). C=A*B, C=A+B, C=A' etc are 2x to 30x faster (10x typical). @JuliaLanguage now has a full-featured interface to GraphBLAS, incl. semirings, masks, etc, for graph algo (@MATLAB uses GrB but only for C=A*B)
SuiteSparse:GraphBLAS v6.1.4 is out with improved documentation and compile-time flags for cpu_features. It's the first version thoroughly tested on Apple Silicon (thanks to @szarnyasg).
https://t.co/cTEEdKOb9J
This paper presents a benchmark for the incremental evaluation of graph queries on the LDBC social network data sets. It includes tools from multiple domains (relational and graph DBMSs, model transformation tools, differential dataflow, GraphBLAS).
In this paper, researchers presented 4 new algorithms for performing parallel masked SpGEMM which, in most cases scenarios, significantly outperform the state of the art for Masked SpGEMM implementations on a large number of real-world benchmarks. #HPC
https://t.co/5m6AZ56Nr6