Document Details
|
Document Type |
: |
Thesis |
|
Document Title |
: |
Big Data and HPC Integration: An Investigation GPUs, Deep Learning and In-Memory Big Data Computing التكامل بين البيانات الكبيرة والحوسبة عالية الاداء: دراسة تشمل وحدات معالجة الرسوم، والتعلم العميق، والذاكرة الداخلية لحوسبة البيانات الكبيرة |
|
Subject |
: |
Faculty of Computing and Information Technology |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
Convergence of big data and high performance computing (HPC) paradigms provides an unimaginable potential for developing new computing paradigms, solving long-standing grand challenges, and making new explorations and discoveries.
The aim of this thesis is to investigate the convergence of big data and HPC technologies using an important application area that is both data and compute-intensive. In this context, this thesis makes the following contributions. Firstly, it proposes a framework for the convergence of big data and HPC using four different types of cutting-edge technologies: big data, in-memory computing, deep learning and Graphical processing units (GPUs). The framework has been implemented using R, TensorFlow, Keras, and Python. The novelty of our approach lies in the integration of the four technologies that are complementary to each other and collectively provide the potential to address big data challenges in a comprehensive manner. The integration of these four technologies has also allowed investigating the viability and benefits of convergence of big data and HPC technologies and paradigms.
Secondly, we apply the proposed framework to four high-impact smart city applications using detailed case studies. These include (1) road traffic speed, flow, and occupancy prediction; (2) road incident prediction; (3) disaster management; and (4) passengers’ entry, exit, and spatio-temporal prediction for London Underground. All four case studies use real open data from sources including California Department of Transportation (Caltrans) Performance Measurement System (PeMS), Transport for London (TfL) Rolling Origin and Destination Survey (RODS), and UK Department for Transport (DfT). Data volume, velocity, variety, veracity, and fusion issues are addressed using over eleven years of traffic data. Different combinations of the datasets along with different network configurations of the deep learning models are investigated for the training and prediction purposes. The results show the usefulness and high-impact of our methods in all four smart city applications.
The convergence framework and the four case studies have contributed multiple novel deep learning models, algorithms, implementation and analytics methodologies, and a range of software products for smart cities, big data, HPC, and their convergence. |
|
Supervisor |
: |
Dr. Ahmed Al-Zahrani |
|
Thesis Type |
: |
Doctorate Thesis |
|
Publishing Year |
: |
1440 AH
2019 AD |
|
Added Date |
: |
Monday, April 29, 2019 |
|
Researchers
| محمد عاقب | Aqib, Mohammad | Researcher | Doctorate | |
|
Back To Researches Page
|