Memory Storage in a Variable Threshold Artificial Neural Network
Abstract
This thesis is concerned with the problem of memory storage in a feedback neural network. A new learning approach has been developed that assigns variable thresholds to the neurons through an iterative process. This approach has been further applied to B-Matrix approach of memory retrieval. Experiments have also been conducted on non-binary neural networks and their storage capacity has been evaluated.
Collections
- OSU Theses [15752]