Optical Computing Research
Abstract
The research focused on understanding the global as well as local properties of the neural network model. Global properties are the dynamics of the network, convergency properties, computational power and capacity. Local properties mean the theory of threshold logic elements, the basic building blocks of the network. Investigated is the relation between error-correcting codes and neural networks. The motivation was that a neural network model can be viewed as a decoder. The stable states correspond to codewords, the probe vector corresponds to the received vector, and convergence to the closest stable state corresponds to Maximum Likelihood Decoding (MLD). Several natural ways were found for connecting the concepts of error correcting codes with the concept of neural networks. The MLD problem in a linear block code is equivalent to finding the global maximum of the energy function of a neural network that can be easily constructed knowing the basis set of the code.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1988
- Accession Number
- ADA202963
Entities
People
- Jehosua Bruck
- Joseph W. Goodman
Organizations
- Stanford University