TY - GEN
T1 - New architecture of correlated weights neural network for global image transformations
AU - Golak, Sławomir
AU - Jama, Anna
AU - Blachnik, Marcin
AU - Wieczorek, Tadeusz
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2018.
PY - 2018
Y1 - 2018
N2 - The paper describes a new extension of the convolutional neural network concept. The developed network, similarly to the CNN, instead of using independent weights for each neuron in the network uses related weights. This results in a small number of parameters optimized in the learning process, and high resistance to overtraining. However unlike the CNN, instead of sharing weights, the network takes advantage of weights correlated with coordinates of a neuron and its inputs, calculated by a dedicated subnet. This solution allows the neural layer of the network to perform global transformation of patterns what was unachievable for convolutional layers. The new network concept has been confirmed by verification of its ability to perform typical image affine transformations such as translation, scaling and rotation.
AB - The paper describes a new extension of the convolutional neural network concept. The developed network, similarly to the CNN, instead of using independent weights for each neuron in the network uses related weights. This results in a small number of parameters optimized in the learning process, and high resistance to overtraining. However unlike the CNN, instead of sharing weights, the network takes advantage of weights correlated with coordinates of a neuron and its inputs, calculated by a dedicated subnet. This solution allows the neural layer of the network to perform global transformation of patterns what was unachievable for convolutional layers. The new network concept has been confirmed by verification of its ability to perform typical image affine transformations such as translation, scaling and rotation.
KW - CNN
KW - Network architecture
KW - Spatial transformation
UR - https://www.scopus.com/pages/publications/85054856264
U2 - 10.1007/978-3-030-01421-6_6
DO - 10.1007/978-3-030-01421-6_6
M3 - Conference contribution
AN - SCOPUS:85054856264
SN - 9783030014209
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 56
EP - 65
BT - Artificial Neural Networks and Machine Learning – ICANN 2018 - 27th International Conference on Artificial Neural Networks, 2018, Proceedings
A2 - Manolopoulos, Yannis
A2 - Hammer, Barbara
A2 - Maglogiannis, Ilias
A2 - Kurkova, Vera
A2 - Iliadis, Lazaros
PB - Springer Verlag
T2 - 27th International Conference on Artificial Neural Networks, ICANN 2018
Y2 - 4 October 2018 through 7 October 2018
ER -