Article · Bulletin of the National Engineering Academy of the Republic of Kazakhstan · December 30, 2024
Comparative analysis of the efficiency of using gpus in machine learning
Maksat Mustafin, Erlan Makhmut
Abstract
In today’s world, data processing and machine learning play a key role in various sectors, including science, industry and technology. As the volume of data increases, the need for effective methods and tools for their analysis increases. Graphics processing units (GPUs) become a necessary element in this process, as they allow complex calculations to be performed in parallel. This article analyzes the efficiency of GPU usage in machine learning. The research covers the use of the RAPIDS libraries (cuMl and cuDF) to perform data analysis and implement machine learning algorithms on the GPU. Alternative machine learning methods, including linear regression and SVM, implemented using the RAPIDS and Scikit-Learn libraries, are also being studied. The presented analysis highlights the advantages of using GPU in training models, especially when working with large amounts of data. The algorithms considered in the study were tested on a dataset created using the Buckley-Leverett mathematical model to predict the oil recovery coefficient. The article demonstrates the importance and potential of using GPUs in machine learning, highlighting key methods that contribute to improving the performance and efficiency of algorithms.
More by Mustafin Maksat
- Adaptive Grid Generation by Solving the Two-Dimensional Diffusion Equation Using Physics-Informed Neural NetworksAlgorithms · 2026
- Efficiency of Parallel Machine Learning Algorithms on Multiprocessor and GPU ArchitecturesLecture notes in computer science · 2025 · Citations: 1
- Adaptive Grid Generation by Solving One-Dimensional Diffusion Equation Using Physics-Informed Neural NetworksAlgorithms · 2025 · Citations: 1
- APPLICATION OF PINN AND THE METHOD OF DIFFERENTIAL CONSTRUCTION OF ADAPTIVE ONE-DIMENSIONAL COMPUTATIONAL GRIDSJournal of problem in computer science and information technologies · 2024 · Citations: 1
