Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

USE OF PARALLEL ALGORITHMS FOR LEARNING AND WORK ARTIFICIAL NEURAL NETWORK

Shirma A.A. 1 Chulyukov V.A. 1
1 Voronezh State Pedagogical University
The usage of parallel computing on multi-core central (CPU) and graphics (GPU) processors to improve performance of an artificial neural network (ANN) in a noise filtering. To some extent, multiprocessor systems copy brain structure. Preliminary results showed 1.5 times performance improvement in image processing and 4-fold at training the ANN. Thus, the use of parallel implementations of algorithms for learning and working ANN can reduce the amount of time required to process the data. For the calculations used CPU Intel Core i5-2400 with 4 cores and graphics card NVIDIA GeForce GTX 460 with 336 cores. As ANN used a multilayer perceptron. Testing was conducted in MATLAB to set Parallel Computing Toolbox.