Title:	Massively parallel computations for piecewise invariants
DNr:	Berzelius-2024-35
Project Type:	LiU Berzelius
Principal Investigator:	Björn Wehlin <bwehlin@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2024-01-22 – 2024-08-01
Classification:	10105
Keywords:

Abstract

In Topological Data Analysis (TDA) we produce features that are piecewise functions (PCFs). These features can then be fed to various Machine Learning algorithms for classification, clustering, and so on. In this project, we have developed a computational tool that translates PCFs to a format that is suitable for machine learning. This tool runs on multi-GPU and multi-CPU configurations. We are in the final stages of this project; the software has been written and tested successfully on 1-2xGPU and 1-18xCPU. We wish to benchmark running the same test suite on 1-8xGPU and a larger CPU.