Parallel Techniques for Image Processing and Artificial Neural Network Simulation
The recent emergence of systems composed of multiple processing elements and memory units, and their associated models of computation promise to alleviate many of the limitations of conventional Von Neumann architectures. The implication of this to the field of Artificial Intelligence is twofold, Parallel systems offer both a significant increase in computing power/speed available, and a more natural physical architecture for implementing parallel solutions to A.I. problems. However, these systems are often extremely complex both from a conceptual (design) and practical (implementation) point of view. In this paper we will analyse various parallel methods and the considerations in using these for problem solving in the areas of image processing and artificial neural network simulation.
Unable to display preview. Download preview PDF.
- Hoare, C.A.R., Ed., “Occam2 Reference Manual”, INMOS Limited, Prentice Hall, 1988.Google Scholar
- Inouchi, H., McLoughlin, N., & Vernon, D., “A real-time simulated human vision system using connectionist models applied to target tracking”, The Proceedings of the 1APR Workshop on Machine Vision Applications, 1990, pp 303–306.Google Scholar
- G.L.Heileman and M.Georgiopoulos. “The Augumented ARTI Neural Network”, Proceedings of the IJCNN,July 1991, volt, pp467–472.Google Scholar
- Iazzetta, A., Vaccaro, R., & Villano, U., “A transputer implementation of boltzman machines”, Parallel architectures and Neural Networks, 1988, pp 128–145.Google Scholar
- Richards, G., Tollenaere, T., “A revised version of the Rhwydwaith neural net simulator”, ECSP-UG-7, 1989.Google Scholar
- Y.Fujimoto, N.Fukuda, “An Enhanced Toroidal Lattice Architecture Neurocomputer for Large Scale Neural Networks”, Proceedings of the IJCNN,June 1989, vollI,, pp 614.Google Scholar
- Clarke, L., Wilson, G., “Tiny: An effecient communications harness for the INMOS transputer”, ECSP-UG-9, 1989.Google Scholar