Loading...

Evaluation of Implementing Machine Learning Methods on Multi-Access Edge Computing Based Networks, Focusing on Computation Offloading

Noorzad, Soroosh | 2023

37 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56584 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Hajsadeghi, Khosrow
  7. Abstract:
  8. Many applications that require heavy processing, such as augmented reality, facial recognition, self-driving cars, and digital healthcare systems, are emerging. Many devices in the Internet of Things space cannot process such a volume of calculations. One of the newest methods to solve this problem is using new architectures in network design, among which we can refer to the edge computing architecture. In this method, by offloading the computation on the available computing resources close to the network's end nodes, we try to provide the results of the calculations for the end nodes of the network as quickly as possible. In this thesis, the problem of determining the computation offloading destination is first investigated using an analytical method. After finding the optimal solution, a data set is provided based on it. Then, using this data set, different machine learning methods were used to solve this problem, and finally, the calculated results were compared with each other. The results of the investigations in this thesis have shown the better performance of machine learning methods from the time delay aspect (an increase of 330 times in the speed of solving the decision-making problem). Finally, some ideas for further work have been presented to improve the accuracy of the methods based on machine learning
  9. Keywords:
  10. Computation Offloading ; Resources Allocation ; Machine Learning ; Deep Learning ; Decision Making Problems Under Uncertainty ; Edge User Allocation ; Multi-Access Edge Computing

 Digital Object List

 Bookmark