Abstract | ||
---|---|---|
A high-order feedforward neural architecture, called pit-sigma (¿t¿) neural network, is proposed for lossy digital image compression and reconstruction problems. The ¿t¿ network architecture is composed of an input layer, a single hidden layer, and an output layer. The hidden layer is composed of classical additive neurons, whereas the output layer is composed of translated multiplicative neurons (¿t-neurons). A two-stage learning algorithm is proposed to adjust the parameters of the ¿t¿ network: first, a genetic algorithm (GA) is used to avoid premature convergence to poor local minima; in the second stage, a conjugate gradient method is used to fine-tune the solution found by GA. Experiments using the Standard Image Database and infrared satellite images show that the proposed ¿t¿ network performs better than classical multilayer perceptron, improving the reconstruction precision (measured by the mean squared error) in about 56%, on average. |
Year | DOI | Venue |
---|---|---|
2007 | 10.1007/s00500-006-0052-z | Soft Comput. |
Keywords | DocType | Volume |
Neural networks,Image compression,Multiplicative neurons,High-order neural networks,Genetic algorithm | Journal | 11 |
Issue | ISSN | Citations |
1 | 1432-7643 | 2 |
PageRank | References | Authors |
0.36 | 5 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Eduardo Masato Iyoda | 1 | 30 | 4.14 |
Takushi Shibata | 2 | 3 | 0.71 |
Hajime Nobuhara | 3 | 192 | 34.02 |
W. Pedrycz | 4 | 13966 | 1005.85 |
Kaoru Hirota | 5 | 1634 | 195.49 |