As part of Data Science Week 2018, Ariful Azad from Lawrence Berkley National Laboratory will present HipMCL – a distributed-memory parallel algorithm for large-scale network clustering. Event Details: Date: Thursday 17 May, 2018 Time: 9.00am Venue: Pawsey Supercomputing Centre HipMCL parallelizes popular Markov Clustering algorithm (MCL) that has been shown to be one of the mostmore…
Date: Thursday 17 May, 2018
Venue: Pawsey Supercomputing Centre
HipMCL parallelizes popular Markov Clustering algorithm (MCL) that has been shown to be one of the most successful and widely used algorithms to cluster sequence similarity or expression networks. Despite its popularity, MCL’s scalability to cluster large datasets still remains a bottleneck due to high running times and memory demands. HipMCL overcomes these challenges by employing novel parallel algorithms for sparse matrix-matrix multiplication, k selection and finding connected components and by utilizing thousands of processors and hundreds of terabytes of memory available in large supercomputers.
HipMCL can cluster large-scale networks 1000 times faster than the original MCL without any information loss. HipMCL can cluster a network with 70 million nodes and 68 billion edges in 2.4 hours using 2000 nodes of a Cray XC40 supercomputer, enabling unprecedented discoveries in network biology. HipMCL is based on MPI and OpenMP and is freely available under a modified BSD license (https://bitbucket.org/azadcse/hipmcl/).
Thu 9:00 am-10:00 am