Improved architecture of the feedforward neural network for image recognition. - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Improved architecture of the feedforward neural network for image recognition.

Résumé

Recently, the researchers have been focusing on the convolutional neural network due to its high reliability in image recognition. It is proposed that the feedforward neural network could compete equivalently with the convolutional neural network? In this paper, we have explored the possibility and proposed a feed-forward neural network, namely the scaled conjugate gradient back-propagation feed-forward neural network with random connections (SCGBP-FNN-RC) to learn big data through recognizing images from the widely known MNIST dataset which is applied with affine and elastic distortions. Based on our findings, SCGBP-FNN-RC has managed to achieve a state-of-the-art accuracy of 99.54 %. The proposed networks are then evaluated based on various parameters, e.g. the neuron size, connections and the Κ parameter.
Fichier non déposé

Dates et versions

hal-04313208 , version 1 (29-11-2023)

Identifiants

Citer

Ching Loong Seow, Mina Aziz, Samer Yahya, Haider A.F. Almurib, Mahmoud Moghavvemi. Improved architecture of the feedforward neural network for image recognition.. 2016 IEEE Industrial Electronics and Applications Conference (IEACon), IEEE, Nov 2016, Kota Kinabalu, Malaysia. pp.280-286, ⟨10.1109/IEACON.2016.8067392⟩. ⟨hal-04313208⟩
1 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More