Search Techniques for Self-Organizing Systems
Abstract
This study has been devoted to extension and further development of search algorithms of utility for self-organizing control systems. The four major results of this study are as follows: Self-organizing search methods developed in the previous study (AMRL-TR-73-76) have been extended to higher- dimensional multi-modal problems and have been shown to be very effective. A composite search algorithm incorporating both the pdf-guided search and the guided accelerated random search was found to be more effective than either search algorithm alone. Clustering analysis has been shown to be a valuable tool for assessing the complexity of a search surface. The number of modes (peaks), their locations relative to each other, their shape and volume, and the estimated maximum performance value within each are all adaptively determined via clustering. A new method for image encoding has been formulated that provides image reconstruction of similar quality to methods currently in use. This procedure also can find regions of possible interest within the image because of its ability to treat the image as a whole rather than line-by-line. This characteristic considerably enhances its value as a tool in image pattern recognition and classification.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1975
- Accession Number
- ADA014815
Entities
People
- Anthony N. Mucciardi
- Elizabeth R. Johnson