Fast Clustering Using Bohrium

A number of scientific instruments needs to cluster observations into a single event, thus hard real-time clustering is required. The proposed project is to build a hardware solution to clustering for GPGPUs in Bohrium. The number of observations and clusters will be reasonably large, but small enough that the system can fit on a large GPGPU, but speed requirements will be very high, typically less than one second.

Area: Project Bachelor Masters

Tags: Bohrium:scientific-programming

Contact: Brian Vinter, or Mads Kristensen,

Activities: Analysis, design and implementation