Conferences  >  Informatics  >  Distributed, Edge and Cloud Computing  >  Germany

The Hausdorff Research Institute for Mathematics (HIM)

The last decade has seen an explosion of available data. To analyze the enormous amount of accrued data, classical algorithms are often not appropriate: the data may not be accessible in its entirety (but perhaps only in an aggregate form) and computations may need to be distributed or done in an online setting. The data often comes with an additional geometric flavor, either from applications or from mapping the data to feature spaces. The workshop will focus on three classes of algorithms for dealing with massive geometric data, namely, streaming, distributed, and sublinear algorithms. The aim of the workshop to identify and explore new research directions at the interface of massive data models and computational geometry, through fruitful discussions leading to potential collaborations.

Schloss Dagstuhl - Leibniz-Zentrum für Informatik

In modern computational systems, the need to handle large-scale inputs imposes interesting computational challenges. Two such challenges are (1) the need to distribute the computation over multiple units, and (2) the dynamic nature of the input, which may undergo changes over time. A particular class of problems studied in these settings is when the input to the computational task is a huge graph. The field of dynamic graph algorithms addresses efficiently processing edge/vertex insertions/deletions in the input graph. In distributed graph algorithms, the input resides across multiple machines, and the goal is to solve the problem while minimizing the number of rounds of communication. Both of these rich research areas have been extensively studied since at least the 1980’s. We know of efficient algorithms for a large variety of tasks, such as shortest paths problems, coloring, subgraph finding, symmetry breaking, approximations, and many more. However, there are still fundamental problems with no known efficient solutions in some of these models, and even more where the exact complexity of computation is yet to be determined. In the recent years, a number of influential works show how transferring ideas from one of these models to the other provides progress on some of the long-lasting open problems. The goal of this Dagstuhl Seminar is to build a bridge between the two research communities of dynamic graph algorithms and distributed computing, by working together on joint research frontiers.

High Performance Computing, Supercomputing, Parallel Computing,    Scientific Computing and Data Science