Method of construction of parallel systems for fuzzy logical inference based on GPU accelerators
We examine a new method for constructing parallel hierarchical systems of fuzzy logic inference using multi-tier parallel form of algorithms based on Nvidia GPU accelerators and CUDA technology. The efficiency of application of hierarchical systems of fuzzy inference for development of diagnostic in...
Saved in:
| Date: | 2018 |
|---|---|
| Main Authors: | , |
| Format: | Article |
| Language: | Ukrainian |
| Published: |
PROBLEMS IN PROGRAMMING
2018
|
| Subjects: | |
| Online Access: | https://pp.isofts.kiev.ua/index.php/ojs1/article/view/306 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Problems in programming |
| Download file: |  |
Institution
Problems in programming| Summary: | We examine a new method for constructing parallel hierarchical systems of fuzzy logic inference using multi-tier parallel form of algorithms based on Nvidia GPU accelerators and CUDA technology. The efficiency of application of hierarchical systems of fuzzy inference for development of diagnostic intelligent system is substantiated. We characterize the organization of efficient computations on graphic accelerators with the aim of achieving maximum degree of parallelism in fuzzy hierarchical systems. In particular, the above organization relies on parallelization of rules inside each block of fuzzy rules. An intelligent software system for assessing the quality of startups based on parallel inference architecture that contains Takagi – Sugeno blocks of fuzzy rules is developed. The experiment is conducted to demonstrate acceleration estimates for developed intellectual system for assessing the quality of startups as well as for more complex systems with randomly generated dependencies between blocks of rules. We compare characteristics of the obtained acceleration estimates with corresponding estimates for hierarchical fuzzy systems based on the distributed computing technology and MPI message exchange. The advantages of developed method for construction of parallel fuzzy inference based on GPU are substantiated. Problems in programming 2017; 4: 003-015 |
|---|