Pregled bibliografske jedinice broj: 719443
GPU Implementation of the Feedforward Neural Network with Modified Levenberg-Marquardt Algorithm
GPU Implementation of the Feedforward Neural Network with Modified Levenberg-Marquardt Algorithm // 2014 International Joint Conference on Neural Networks (IJCNN)
Peking: Institute of Electrical and Electronics Engineers (IEEE), 2014. str. 785-791 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 719443 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
GPU Implementation of the Feedforward Neural Network with Modified Levenberg-Marquardt Algorithm
Autori
Bacek Tomislav, Dubravko Majetić, Danko Brezak
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
2014 International Joint Conference on Neural Networks (IJCNN)
/ - Peking : Institute of Electrical and Electronics Engineers (IEEE), 2014, 785-791
ISBN
978-1-4799-6627-1
Skup
2014 International Joint Conference on Neural Networks (IJCNN)
Mjesto i datum
Peking, Kina, 06.07.2014. - 11.07.2014
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
Feedforward Neural Network; Levenberg-Marquardt Algorithm
Sažetak
In this paper, an improved Levenberg-Marquardt based feedforward neural network, with variable weight decay, is suggested. Furthermore, parallel implementation of the network on graphics processing unit is presented. Parallelization of the network is achieved on two different levels. First level of parallelism is data set level, where parallelization is possible due to inherently parallel structure of the feedforward neural networks. Second level of parallelism is Jacobian computation level. Third level of parallelism, i.e. parallelization of optimization search steps, is not implemented due to the variable weight decay, which makes third level of parallelism redundant. Suggested weight decay variation enables the compromise between higher accuracy with oscillations on one side and stable, but slower convergence on the other. To improve learning speed and efficiency, modification of random weight initialization is included. Testing of proposed algorithm is performed on two real domain benchmark problems. The results obtained and presented in this paper show effectiveness of proposed algorithm implementation.
Izvorni jezik
Engleski
Znanstvena područja
Strojarstvo
POVEZANOST RADA
Projekti:
120-1201842-3048 - Umjetna inteligencija u upravljanju složenim nelinearnim dinamičkim sustavima (Kasać, Josip) ( CroRIS)
120-1201948-1945 - Inteligentno vođenje obradnih sustava (Majetić, Dubravko, MZOS ) ( CroRIS)
120-1201948-1938 - Napredni obradni sustavi i procesi (Udiljak, Toma, MZOS ) ( CroRIS)
Ustanove:
Fakultet strojarstva i brodogradnje, Zagreb