In this article, we formulate the image reconstruction
In the early 1980s, Hopfield 11] presented a model of problem in terms of a multicriteria optimization-based neural network neural computation that was based on the interaction of neurons.
model, and study its performance. The value of neural network as a
The model consisted of a set of first-order nonlinear differential computational device is to be shown. We also demonstrate the effiequations that minimize a certain ''energy'' function. He argued ciency of our implementation of neural net processing and compare that there are emergent computational capabilities at the network its performance with respect to the convolution-backprojection relevel that are nonexistent at the single neuron level. This kind of construction techniques based on computer-generated noisy projecneural networks is now known as a Hopfield net.
tions. The method presented here has the particular ability to solve
In the mid 1980s, Rumelhart and colleagues [12] published ill-posed reconstruction problems. The algorithm described was imtheir landmark books on parallel distributed processing, establishplemented on a serial PC/Intel 586 microcomputer, but was cast in a form which is ideally suited for parallel processing. แญง 1998