October 7th through 10th Computing Ph.D. student Matthew Merris attended the Chesapeake Large Scale Analytics Conference (CLSAC) in Annapolis, Maryland. This year’s conference explored “efforts to address the societal impacts of large-scale analytics.” Merris presented a poster on his current work ‘Advancing the Bases: Tensor Methods and The Frontiers of Generality.’
CLSAC “is a conference organized by the Association for High-Speed Computing that brings together experts from industry, academic, and government institutions to discuss current and future challenges for data analytics.”
Merris’s poster on his current work, named ‘The Tensor500 project,’ is “exploring what constitutes sound benchmarking and metrics for high-performance systems by leveraging tensors and tensor methods to craft a series of composable metakernels.” Merris went on to summarize his project stating, “It is our belief that allowing end-users to build representative workflows will provide a robust and actionable assessment of implementation. A tensor is an array of arbitrary dimensions (i.e. a vector is a 1-dimensional tensor, a matrix is a 2-dimensional tensor, etc). This generality makes the tensor well suited and ubiquitous at the forefront of large-scale, high-dimensional streaming data analytics. Current benchmarks and metrics provide nominal feedback on the performance and suitability of the increasingly heterogeneous systems tasked with contemporary data-intensive, time-sensitive workloads.”
Merris is a Computing Ph.D. student in the Computer Science Emphasis who is involved with the Tensor500 research group developing benchmarks around tensor-based operations.