High-Performance Computing for Big Data Analytics
My name is Shirin Tavara and I'm a PhD student within IPSI. The current title of my research is: High-Performance Computing for Big Data Analytics. The analysis of big data is commonly utilizing support vector machines, and therefore a large focus of my research is on parallelization of related data mining algorithms and implementation of efficient tools for data analysis. I started my research in the end of 2012 and plan on being finished before the end of 2017. This work is mainly funded by University of Borås and will be conducted in close cooperation with them as well as University of Skövde as found feasible.
This PhD project is part of ongoing research collaboration with the University of Skövde, which aims to strengthen research capacity at both universities. Skills needed in this project are a solid foundation in mathematics and computer science, combined with very good practical skills in programming and systems development. Special weight is previous experience in optimization and parallelization of programs. The development of efficient parallel algorithms is a prerequisite for achieving the objectives of high-performance computing. The availability of high performance computing increases and methods of their use refined. This results in both new opportunities and needs of research in adjacent areas of computer science such as artificial intelligence and data analysis.
This project recognizes that online analysis and prediction of very large data sets require efficient model generation and prediction execution. Luckily, ensembles lend themselves to parallelization, allowing ensemble members to be generated and applied concurrently. As a matter of fact, with the emerging change in hardware with multi-core and many-core architectures available as standard computer equipment, it has become increasingly interesting and important to design parallel versions of earlier sequential algorithms for data mining. In addition to allowing the same tasks as earlier being solved much faster, this enables performing big data analytics that would require prohibitively long time using a standard sequential implementation.
Supervisors and Mentors
People directly involved in my research are:
- Principal Supervisor – Associate Professor Håkan Sundell, University of Borås
- Primary Supervisor – Associate Professor Håkan Sundell, University of Borås
- Assistant Supervisor – Dr. Anders Gidenstam, University of Borås
- Assistant Supervisor – Dr. Alexander Karlsson, University of Skövde
- Assistant Supervisor – Dr. Anders Dahlbom, University of Skövde
- Board of directors of IPSI