Title :
Image coding/reconstruction and matching using a parallel distributed Hebbian architecture
Author_Institution :
Dept. of Syst. Eng. & Eng. Manage., Chinese Univ. of Hong Kong, Shatin, China
fDate :
6/24/1905 12:00:00 AM
Abstract :
A high-gain, post-annealing, generalized Hebbian algorithm is proposed and observed to have chaotic learning behavior. It lends itself readily to a highly efficient parallel distributed architecture for principal components computation. The work is extended to a convergence accelerator that uses the chaotic pattern learned during the first few epochs for an iterative weight-change procedure. Applications of using the parallel architecture for image encoding, reconstruction, and matching are described. Successful simulation results in yielding good quality reconstructed images and photo/sketch matching are reported
Keywords :
Hebbian learning; chaos; image coding; image matching; image reconstruction; neural net architecture; neural nets; parallel architectures; chaotic learning behavior; high-gain post-annealing generalized Hebbian algorithm; image coding; image matching; image reconstruction; iterative weight-change procedure; parallel distributed Hebbian architecture; photo matching; sketch matching; Chaos; Computational modeling; Computer architecture; Concurrent computing; Convergence; Distributed computing; Image coding; Image reconstruction; Iterative algorithms; Parallel architectures;
Conference_Titel :
Neural Networks, 2002. IJCNN '02. Proceedings of the 2002 International Joint Conference on
Conference_Location :
Honolulu, HI
Print_ISBN :
0-7803-7278-6
DOI :
10.1109/IJCNN.2002.1005607
