Abstract
In this paper, we describe implementation of ANN training process using backpropagation learning algorithm for exploiting the high performance SIMD architecture of GPU using CUDA. We also compare sequential and parallel algorithm execution times and conducted speedup analysis for both the methods. The simulation results demonstrate a significant decrease on executing times and greater speedup than serial implementation of training and learning processes. All due to the parallel algorithm and use of the GPU, the training time for huge set of images get reduced significantly increasing the accuracy rate of face recognition.
Original language | English |
---|---|
Pages (from-to) | 801-808 |
Number of pages | 8 |
Journal | Far East Journal of Electronics and Communications |
Volume | 16 |
Issue number | 4 |
DOIs | |
Publication status | Published - 01 Dec 2016 |
Fingerprint
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Electrical and Electronic Engineering
Cite this
}
Parallel backpropagation neural network training for face recognition. / Omarov, Batyrkhan; Suliman, Azizah; Tsoy, Anton.
In: Far East Journal of Electronics and Communications, Vol. 16, No. 4, 01.12.2016, p. 801-808.Research output: Contribution to journal › Article
TY - JOUR
T1 - Parallel backpropagation neural network training for face recognition
AU - Omarov, Batyrkhan
AU - Suliman, Azizah
AU - Tsoy, Anton
PY - 2016/12/1
Y1 - 2016/12/1
N2 - In this paper, we describe implementation of ANN training process using backpropagation learning algorithm for exploiting the high performance SIMD architecture of GPU using CUDA. We also compare sequential and parallel algorithm execution times and conducted speedup analysis for both the methods. The simulation results demonstrate a significant decrease on executing times and greater speedup than serial implementation of training and learning processes. All due to the parallel algorithm and use of the GPU, the training time for huge set of images get reduced significantly increasing the accuracy rate of face recognition.
AB - In this paper, we describe implementation of ANN training process using backpropagation learning algorithm for exploiting the high performance SIMD architecture of GPU using CUDA. We also compare sequential and parallel algorithm execution times and conducted speedup analysis for both the methods. The simulation results demonstrate a significant decrease on executing times and greater speedup than serial implementation of training and learning processes. All due to the parallel algorithm and use of the GPU, the training time for huge set of images get reduced significantly increasing the accuracy rate of face recognition.
UR - http://www.scopus.com/inward/record.url?scp=85006333151&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85006333151&partnerID=8YFLogxK
U2 - 10.17654/EC016040801
DO - 10.17654/EC016040801
M3 - Article
AN - SCOPUS:85006333151
VL - 16
SP - 801
EP - 808
JO - Far East Journal of Electronics and Communications
JF - Far East Journal of Electronics and Communications
SN - 0973-7006
IS - 4
ER -