Using CUDA GPU to Improve Parallelism of OHD-SVM
Autor: | LIN, TZU-HENG, 林子恆 |
---|---|
Rok vydání: | 2019 |
Druh dokumentu: | 學位論文 ; thesis |
Popis: | 107 In recent years, artificial intelligence is one of the popular researches. Due to the powerful computing power of the graphics processing unit, the growth of artificial intelligence is accelerated, and machine learning is a branch of artificial intelligence. Among these machine learning, the support vector machine is a very important area of research. Support vector machine is a kind of supervised learning, while supervised learning can be divided into: binary classification, multi-class classification, and regression. However, support vector machine can be widely used, and it is not limited to a certain classification. With the rise of parallel computing in CUDA, it has been proposed that the support vector machine using CUDA parallelization is called OHD-SVM, and we find that the step of the vector mapping used in the OHD-SVM is a bottleneck on the overall parallelism. Therefore, we propose different parallelization technology of the vector mapping to replace the original method to improve the performance of OHD-SVM. |
Databáze: | Networked Digital Library of Theses & Dissertations |
Externí odkaz: |