Event Entry
What and Who
New Techniques for Designing Fast Parallel Algorithms
Date, Time and Location
Abstract
Web) makes processing them on a single machine a sheer impossibility.
Nowadays, computation is routinely distributed across many machines by
resorting to massively parallel computation frameworks, such as
MapReduce, Hadoop, Spark, or Flume. This motivates a need to design fast
scalable graph algorithms.
In this talk, I will discuss two recently developed techniques for
designing large scale graph algorithms. I will explain how those
techniques yield exponentially faster parallel methods for approximating
maximum matchings. Instantiations of the same techniques also lead to
exponential speed-up for a number of other graph problems, including
approximate vertex cover, maximal independent set, maximal matching and
graph clustering.
Further, I will briefly outline limitations of existing techniques and
discuss research directions suggested by those limitations. I will
conclude by turning to edge computing, a rising cutting-edge model of
computation, and highlight several research challenges inspired by
modern technologies.
https://cs-uni-saarland-de.zoom.us/j/96884270224?pwd=MVpPYzZ4L1ZtMzFMRmtmWDJCODBuZz09
Contact